Content uploaded by H. Sayoud
Author content
All content in this area was uploaded by H. Sayoud on Feb 13, 2025
Content may be subject to copyright.
Investigation on the
Author’s Style and
the Authenticity
of the Holy Quran
3rd Edition - February 2025
Pr Dr Halim Sayoud
Investigation on the Author’s
Style and the Authenticity
of the Holy Quran
by
Pr, Dr. Halim Sayoud
Sayoud.net
Copyright Sayoud.net 2021-2025, All rights reserved.
Note: This book is made under the OPEN ACCESS rules and policies, with no commercial use.
The book is published under the terms of a Creative Commons license, which permits use and
distribution, but no commercial use.
Book Edited by Pr Halim Sayoud
Publisher EDT - SCHOLARPAGE
© Copyright Sayoud.net, 2025, All rights reserved
DOI: 10.5281/zenodo.14865663
February 2025
Abstract
In this book (3rd edition), the author tries to see whether the Quran could be a simple invention of the
Prophet (i.e. written by the Prophet) or really a book from God (i.e. a divine book sent down by Allah)
as claimed in the Islamic religion, in a scientific manner.
The book can be considered as a pure scientific investigation without any form of theological or
ideological point of view. Also, the author does not discuss his personal beliefs on the subject, but only
what the clear results of this investigation show.
The author describes the 19 series of scientific experiments, which were conducted during this
investigation, as well as the new scientific knowledge recently discovered in the Quran. He also
describes several recent scientific research works confirming the authenticity of the holy Book.
All the results reported in those experiments show that t
The Quran could not be written or invented by the Prophet.
Finally, as reported in this book, the scientific knowledge that is discovered in the Quran, and
commented by several famous researchers in the field, confirms this conclusion and shows that the holy
book could not be written by a human being.
About the Author
Dr. Halim Sayoud is full Professor at USTHB University and head of the EDT research team. Although,
he is specialized in Electronics and Computer Sciences, his research works focus on a variety of research
fields such as Natural language processing, Artificial intelligence and Biometrics.
He is the author and co-author of more than 100 scientific articles in conference proceedings, book
chapters, scientific journals, etc. He is also the Editor-in-chief of a scientific journal. His official website
is http://sayoud.net/
© Copyright 2021-2025 Sayoud.net. All Rights Reserved. II
Contents with hyperlinks
1 Introduction ............................................................................................................................................. 1
2 Description of the Holy Quran ................................................................................................................ 7
3 Description of the Hadith ........................................................................................................................ 8
4 Who was the Author of the Quran? ......................................................................................................... 9
5 Stylometric Analysis and Automatic Authorship Attribution ............................................................... 10
6 The Quran and Hadith Corpora ............................................................................................................. 12
6.1 Dimension of the two religious books .......................................................................................... 12
6.2 Segmentation ................................................................................................................................ 12
6.3 Word structure of the different segments ..................................................................................... 13
7 First Series of Experiments: Global Analysis ....................................................................................... 15
7.1 Authorship Discrimination based on Words ................................................................................. 15
7.2 Authorship Discrimination based on Characters .......................................................................... 16
7.3 Authorship Analysis based on COST parameter .......................................................................... 17
7.4 Word length frequency based analysis ......................................................................................... 20
7.5 Discriminative words based analysis ............................................................................................ 21
7.6 Numbers citation based analysis ................................................................................................... 23
7.7 Animal citation based analysis ..................................................................................................... 24
7.8 Special Ending bigrams ................................................................................................................ 26
7.9 Discussion on the first series of experiments................................................................................ 28
8 Second series of experiments: Big Segments based Segmental analysis .............................................. 29
8.1 Discriminative words .................................................................................................................... 29
8.2 Word length frequency based analysis ......................................................................................... 30
COST parameter ........................................................................................................................... 31
Discriminative characters ............................................................................................................. 32
Vocabulary based similarity ......................................................................................................... 32
8.6 Discussion on the second series of experiments ........................................................................... 34
9 Thirdseries of experiments: Automatic authorship attribution with several features and several
classifiers ....................................................................................................................................................... 35
9.1 First experiment ............................................................................................................................ 36
9.2 Second experiment ....................................................................................................................... 36
9.3 Discussion on the third series of experiments .............................................................................. 37
10 Fourthseries of experiments: Short Segments based Segmental Authorship Attribution ............... 38
10.1 Experiments of Authorship Attribution using a Hierarchical clustering ...................................... 39
10.2 Experiments of Authorship Attribution using different state of the art features and classifiers ... 40
10.2.1. Brief description of the different classifiers ................................................................................. 41
10.2.2. Authorship Attribution Results ..................................................................................................... 43
10.3 Experiments of Authorship Attribution using the COST parameter ............................................ 44
10.4 .......................... 45
10.5 Discussion on the fourth series of experiment .............................................................................. 46
11 Fifth Series of Experiment: Stylometric Comparison between the Quran and Hadith based on
Successive Function Words .......................................................................................................................... 47
11.1 Used function words ..................................................................................................................... 47
11.2 Analysis of the books based on the SFW ..................................................................................... 48
11.2.1. Experiment 1 - Comparison between the Quran and Hadith ........................................................ 48
11.2.2. Experiment 2 - Comparison of seven different books: Quran, Hadith and five other books ........ 49
11.3 Discussion on this investigation ................................................................................................... 50
12 Sixth Series of Experiment: Authorship Identification of 7 Books A Fusion Approach ............... 51
12.1 Corpus of the seven religious books ............................................................................................. 51
12.2 Authorship Attribution Methods .................................................................................................. 53
12.2.1 Conventional Classifiers ...................................................................................................... 53
12.2.2 The Fusion approach ............................................................................................................ 54
12.3 Experiments of authorship attribution .......................................................................................... 56
12.3.1 Experiments of authorship attribution using conventional features and classifiers.............. 57
12.3.2 Experiments of authorship attribution using fusion techniques ........................................... 59
12.4 Discussion and comments ............................................................................................................ 60
13 Seventh Series of Experiments: Authorship Discrimination using the Leave-One-Out Validation . 61
13.1 The Leave-One-Out Method ........................................................................................................ 61
13.2 On the Choice of the Classifier and Feature ................................................................................. 65
13.3 Text Segmentation ........................................................................................................................ 65
13.4 Experiments of AA using the LOO and LTO techniques ............................................................. 65
13.5 Discussion..................................................................................................................................... 67
14 Eighth Series of Experiments: Authorship Discrimination based on Gaussianity and Interpolability
68
14.1 Introduction .................................................................................................................................. 68
14.2 Fitting and interpolation definitions ............................................................................................. 69
14.3 Investigation on the Word-Length frequency ............................................................................... 70
14.3.1. Definition of the Word Length Frequency (WLF) ....................................................................... 70
14.3.2. Graphical representation of the Word Length Frequency............................................................. 70
14.3.3. Hadith model interpolated with Gaussian fitting .......................................................................... 76
14.4 Investigation on Numbers citation frequency ............................................................................... 77
14.5 Conclusion and Discussion ........................................................................................................... 81
15 Ninth Series of Experiments: A Mysterious Numerical Structure in the Quran making it different
from other Human Books .............................................................................................................................. 83
15.1 Motivation based on the citation of number 7 in the holy Quran ................................................. 83
15.2 Citation of number 7 in the holy Quran ........................................................................................ 84
15.3 New Statistical Evidences based on Number Seven ..................................................................... 85
15.4 Discussion..................................................................................................................................... 90
16 Tenth Series of Experiments: Authorship Attribution based on the Interrogative Form .................. 91
16.1 Text Segmentation ........................................................................................................................ 91
16.2 Proposed Interrogative Features ................................................................................................... 91
16.3 Authorship Discrimination Approach ........................................................................................... 94
16.4 Feature extraction and LFF fusion ................................................................................................ 95
16.5 Classification methods .................................................................................................................. 96
16.6 Experimental results and analysis ................................................................................................. 96
16.7 Discussion..................................................................................................................................... 98
17 Eleventh Series of Experiments: Investigation on the Quran/Hadith Authorship Using Visual
Analytics Approaches ................................................................................................................................... 99
17.1 Stylometric features ...................................................................................................................... 99
17.1.1. ............................................................................................... 99
17.1.2. ...................................................................... 100
17.1.3. Frequency of some discriminative words ................................................................................... 100
17.1.4. On the COST parameter ............................................................................................................. 100
17.1.5. Word length frequency ............................................................................................................... 100
17.1.6. Frequency of the coordination conjunction « » (meaning AND in English) ........................... 100
17.1.7. Frequency of the coordination conjunction Waw « » at the beginning of sentence................. 101
17.2 Visual Analytics based Clustering methods ............................................................................... 101
17.2.1 Hierarchical clustering ....................................................................................................... 102
17.2.2 C-Means clustering ............................................................................................................ 103
17.2.3 K-Means clustering ............................................................................................................ 104
17.2.4 Sammon Mapping .............................................................................................................. 105
17.2.5 Principal Components Analysis ......................................................................................... 107
17.2.6 Gaussian Mixture Model based clustering ......................................................................... 108
17.2.7 Self-Organizing Map based clustering ............................................................................... 111
17.3 Results summarization ................................................................................................................ 113
17.4 Discussion................................................................................................................................... 114
18 Twelfth Series of Experiments: Authorship Discrimination based on Word Transition Probability
116
18.1 Probability Computation Procedure ........................................................................................... 116
18.2 Probability Normalization Procedure ......................................................................................... 117
18.3 Selection of the bigrams ............................................................................................................. 117
18.3.1 Case of limited set of bigrams ............................................................................................ 117
18.3.2 Case of unlimited set of bigrams ........................................................................................ 118
18.4 Experimental Results .................................................................................................................. 118
18.4.1 Experiments with limited bigrams ..................................................................................... 118
18.4.2 Experiments with unlimited bigrams ................................................................................. 120
18.5 Discussion and Conclusion ......................................................................................................... 121
19 Thirteenth Series of Experiments: Authorship Discrimination based on Deep Learning Technology
122
19.1 Dataset ........................................................................................................................................ 122
19.2 Proposed Model based on LSTM ............................................................................................... 123
19.3 Experimental Results .................................................................................................................. 124
19.4 Conclusion .................................................................................................................................. 125
20 Book Analysis based on Embedded Scientific Knowledge ............................................................ 126
20.1 Scientists talking about the scientific aspect of the Quran ......................................................... 126
20.2 Number of Months and Days in the Quran: "An Enigma" ......................................................... 129
20.3 Earth Rotation in the Quran ........................................................................................................ 129
20.4 Expansion of the Universe in the Quran ..................................................................................... 131
20.5 A Scientific Evidence on the Sun Movement in the Holy Quran ............................................... 132
20.6 About the Embryo description in the Quran ............................................................................... 133
20.7 Description of the Pharaoh's death and the preservation of his body in the Quran .................... 134
21
Religious Books .......................................................................................................................................... 136
21.1 Last Prophet Concept and Truthfulness of the Holy Quran ........................................................ 136
21.2 Other significations on the fact that there will be no more Prophets .......................................... 137
21.3 Prediction of Muhammad in the Bible and ancient religious books ........................................... 138
21.4 Conclusion .................................................................................................................................. 140
22 Does the Heart have a Control on Mind and Emotions? A Scientific Evidence Supporting what is
Said in the Holy Quran ................................................................................................................................ 141
22.1 What is the main source of the mind and emotions: the Heart, the Brain or both? .................... 141
22.2 Heart and Brain citation in the Quran ......................................................................................... 142
22.3 Conclusion .................................................................................................................................. 143
23 Do Animals communicate with each other? A Scientific Evidence Supporting what was Revealed
in the Quran ................................................................................................................................................. 144
23.1 Animal communication in the Quran ......................................................................................... 144
23.2 Scientific discoveries about animal communication .................................................................. 145
23.2.1 Dolphins communication ................................................................................................... 146
23.2.2 Bird communication ........................................................................................................... 147
23.2.3 Ant communication ............................................................................................................ 147
23.2.4 Bee communication ............................................................................................................ 148
23.3 Conclusion .................................................................................................................................. 148
24 Effect of the holy Quran in Soul Appeasement and Treatment of Anxiety: An experimental
Evidence on the Divinity of the Book ......................................................................................................... 150
24.1 Experimental studies................................................................................................................... 150
24.1.1 Study 1 ............................................................................................................................... 150
24.1.2 Study 2 ............................................................................................................................... 151
24.1.3 Study 3 ............................................................................................................................... 152
24.1.4 Study 4 ............................................................................................................................... 152
24.1.5 Summary of the four studies .............................................................................................. 152
24.2 Quran effect in coping with anxiety ........................................................................................... 153
24.3 Discussion................................................................................................................................... 154
25 Statistical Investigation on Ancient Quran Folios: Case of Birmingham and Sanaa Parchments .. 155
25.1 Introduction ................................................................................................................................ 155
25.1.1 Introduction on the Birmingham Quran manuscript .......................................................... 155
25.1.2 Introduction on the Sanaa Quran manuscript ..................................................................... 156
25.1.3 Goal of this work ................................................................................................................ 156
25.2 Notes on the ancient Arabic handwriting ................................................................................... 156
25.2.1 Hijazi script ........................................................................................................................ 156
25.2.2 Diacritics and rasm dots ..................................................................................................... 157
25.2.3 Note on the "silent alif" ...................................................................................................... 157
25.3 Analysis of the Birmingham Quran ............................................................................................ 157
25.4 Analysis of the Sanaa Quran ...................................................................................................... 159
25.4.1 Statistical analysis of the Sanaa folio ................................................................................. 159
25.4.2 Comparison between Sanaa and Birmingham folios .......................................................... 159
25.5 Discussion................................................................................................................................... 160
26 General Conclusion and Discussion ............................................................................................... 162
27 Personal Feeling.............................................................................................................................. 165
References ................................................................................................................................................... 166
© Copyright Sayoud.net 2021-2025, All rights reserved. 1
1 Introduction
In the Islamic religion, it is well known that the holy Quran is the book of Allah (God) and is claimed
to be written (i.e. created) by him. So the Author of the holy Quran is known to be Allah (God), as
stated in the holy Quran and confirmed by the Prophet (pbuh).
For instance, the following Quran verses clearly confirm that authorship.
Verse
Ref.
4:82
Translation: Then do they not reflect upon the Qur'an? If
it had been from [any] other than Allah, they would have
found within it much contradiction.
Verse
Ref.
10:37
Translation: And it was not [possible] for this Qur'an to
be produced by other than Allah, but [it is] a confirmation
of what was before it and a detailed explanation of the
[former] Scripture, about which there is no doubt, from
the Lord of the worlds.
Verse
Ref.
12:3
Translation: We narrate to you the best of narratives, by
Our revealing to you this Quran, though before this you
were certainly one of those who did not know.
Verse
Ref.
27:6
Translation: Indeed you receive the Quran from One
who is all-wise, all-knowing.
Verse
Ref.
17:88
Translation: Say: "If all mankind and all invisible beings
would come together with a view to producing the like of
this Qur'an, they could not produce its like even though
they were to exert all their strength in aiding one
another!"
© Copyright Sayoud.net 2021-2025, All rights reserved. 2
Verse
Ref.
76:23
Translation: Indeed, it is We who have sent down to
you, [O Muhammad], the Qur'an progressively.
Moreover, the Prophet (Pbuh) clearly confirmed that the Quran was sent down to him by God, as
expressed in the following Hadith statements − the Hadith represents the statements and speech of the
Prophet.
Hadith 1
Translation: Surah Al-
Anaam (chapter 6 of the Quran) was sent down to him at
Hadith 2
" .
Translation:
down to me, which I love more than this world
Hadith 3
:
Translation: The Prophet Pbuh said to his companion
recite the Quran in
Hadith 4
Translation: The Prophet Pbuh said:
matters with you. As long as you hold to them, you will
not go the wrong way. They are the Book of Allah
(Quran) and the Sunna of His Prophet (Hadith)."
© Copyright Sayoud.net 2021-2025, All rights reserved. 3
So, in all those statements from the Hadith, it is clear that the Prophet confirmed that the Quran was
sent down to him from Allah (God).
Some unstructured doubts on the origin of the Quran
Some ignorant persons tried over the time to claim that the holy Quran is only an invention of the
Prophet and that it could be written by him, instead of God, without providing any proof.
Strangely, one can retrieve such claims even in the Quran when describing some ignorant or malicious
persons describing the holy Quran as a simple invention of the Prophet without any link to God
(Allah), as stated in the following verses (32-3):
Translation: Or do they say, "He invented it"? Rather, it is the truth from your Lord, [O Muhammad],
that you may warn a people to whom no warner has come before you [so] perhaps they will be guided.
Purpose of this research work
Due to some doubts on the origin of the holy Quran, from certain untrustworthy persons, it becomes
necessary to undertake a scientific investigation on the matter to try discarding the doubts or at least
clarifying the problem.
The solution, in the opinion of the author, is a clear scientific investigation that should shed the light
on the problem and hopefully provides a final answer to all related questions.
In the age of computer sciences and artificial intelligence, it is commonly becoming possible to tackle
some problems that were quite difficult to handle before, like biometrics, for instance, which has been
widely used nowadays to recognize human beings in different situations. So, the key methodology
should focuses on such intelligent techniques with maybe some text-mining tools to try providing a
fair scientific answer to the authorship problem.
How to identify an author?
Human beings have distinctive ways of speaking and writing, as explained by Corney (Corney, 2003),
and there exists a long history of linguistic and stylistic investigation into authorship attribution
(Holmes, 1998). In recent years, practical applications of authorship attribution have grown in areas
such as intelligence (linking intercepted messages to each other and to known terrorists), criminal law
(identifying writers of ransom notes and harassing letters), civil law (copyright and estate disputes),
and computer security (tracking authors of computer virus source code). As reported by Madigan
(Madigan et al., 2005), this activity is part of a broader growth within computer science of
identification technologies, including biometrics (retinal scanning, speaker recognition, etc.),
cryptographic signatures, intrusion detection systems, and others.
© Copyright Sayoud.net 2021-2025, All rights reserved. 4
The research field dealing with author recognition This research domain
includes several specialties, such as: Authorship Attribution, Authorship Verification, Plagiarism
Detection, Author Discrimination, Text Segmentation and so on.
There are several features that can be used: characters, vocabulary, sentences, mistakes, character N-
grams, word N-grams, etc. Also, several classification techniques can be employed: distance metrics,
statistics, automatic learning classifiers, neural networks, deep learning and so on.
Is stylometry efficient?
Like many biometric modalities, stylometry can be quite accurate provided that the following
conditions are well respected.
Firstly, the amount of text data should be sufficiently large: a minimum of 2500 words per document
should provide good performances as reported by Eder (Eder, 2010).
Secondly, to be fair, the examined texts should have the same topic and same genre.
Thirdly, the compared text documents should belong to the same period of time and should be written
with the same character type.
If all those conditions are respected, the author recognition performances can reach a quite high level of
accuracy. For instance, during an experiment of authorship attribution on a closed-set dataset of 100
authors, called HAT Corpus (HAT is composed of 100 groups of Arabic texts that are extracted from
100 different Arabic books. The books were written by 100 different authors with the same topic:
Travel), we got an authorship identification accuracy of 97%, while the text documents size was only
1100 words per document. This result and many others conducted by other researchers confirm the
efficiency of stylometry if the previous conditions are respected.
Tackling the religious enigma
In this research work, we deal with a religious enigma, which has not been solved for fifteen hundred
years, as cited by Sayoud in (S. 2010] H. Sayoud, 2010). In fact, as mentioned previously, certain doubts
on the origins of the Holy Quran do exist and some ignorant persons thought that the Holy Quran could
be an invention of the prophet Muhammad, for three purposes (Al-Shreef, 2009):
To facilitate his domination over his followers;
To frighten the unbelievers and those who disobey his orders;
To permit his pleasures.
Several theologians, over time, tried to prove that this assumption was false. They were relatively logical
and clever, but their proofs were not so convincing for many people, due to a lack in scientific rigor.
Similarly, for the Christian religion, there exist several disputes about the origin of some texts of the
Bible. Such disputes are very difficult to solve due to the delicacy of the problem, the religious
sensitivity and because the texts were written a long time ago.
© Copyright Sayoud.net 2021-2025, All rights reserved. 5
Hence, it can be seen why Holmes (Mills, 2003) pinpointed that the area of stylistic analysis is the main
contribution of statistics to religious studies. For example, early in the nineteenth century,
Schleiermacher disputed the authorship of the Pauline Pastoral Epistle 1 Timothy (Mills, 2003). As a
result, other German speaking theologians, namely, F.C. Baur and H.J. Holtzmann, initiated similar
studies of New Testament books (Mills, 2003).
In such problems, it is crucial to use rigorous scientific tools and it is important to interpret the results
very carefully.
Hence, knowing that authors possess specific stylistic features that make them differentiable (Li et al.,
2006), we tried to make some experiments of author discrimination between the Quran and some
shed the light on this enigma.
For this purpose, several investigations, experiments and definitions are presented and commented as
follow:
In chapter 2, we give a brief description of the Holy Quran.
In chapter 3, we give a brief description of the Hadith.
In chapter 5, we define the task of Stylometry and Automatic Authorship Attribution
In chapter 6, we present the Quran and Hadith Corpora
In chapter 7, we describe our first Series of Experiments: A Global Analysis
In chapter 8, we describe our second series of experiments: Big Segments based Segmental analysis
In chapter 9, we explain the details of the thirdseries of experiments: Automatic authorship
attribution with several features and several classifiers
In chapter 10, we describe the fourthseries of experiments: Short Segments based Segmental
Authorship Attribution
In chapter 11, we describe the fifth Series of Experiment: Stylometric Comparison between the Quran
and Hadith based on Successive Function Words
In chapter 12, we introduce and comment the sixth Series of Experiment: Authorship Identification of
7 Books A Fusion Approach
In chapter 13, we present the seventh series of experiments, where we propose the authorship
discrimination method using the Leave-One-Out Validation.
In chapter 14, we describe the authorship discrimination based on Gaussianity and Interpolability
In chapter 15, we talk about a mysterious numerical structure found in the Quran, which makes it
different from other Human Books
In chapter 16, we propose a new method of authorship attribution based on the Interrogative Form
© Copyright Sayoud.net 2021-2025, All rights reserved. 6
In chapter 17, we present an investigation on the Quran/Hadith Authorship Using Visual Analytics
Approaches
In chapter 18, we propose a new method of authorship discrimination based on Word Transition
Probability
In chapter 19, we explore some Embedded Scientific Knowledge in the holy book
In chapter 20, we give a general conclusion
Finally some references are given at the end of the book.
© Copyright Sayoud.net 2021-2025, All rights reserved. 7
2 Description of the Holy Quran
The Quran (in Arabic: al-
Koran, Alcoran or Al-(Wiki1, 2012, p. 1) (Nasr, 2004) is the central religious text of Islam.
Muslims believe the Quran to be the book of divine guidance and direction for mankind (I. A. Ibrahim,
1997) (that has been written by God), and consider this Arabic book to be the final revelation of God.
Islam holds that the Quran was written by Allah (ie. God) and transmitted to Muhammad by the angel
Gibraele (Gabriel) over a period of 23 years. The beginning of Quran apparition was in the year 610
(after the birth of Christ).
Figure 2.1 Very old copy Quran (Image credit: Cadbury Research Library, University of Birmingham)
© Copyright Sayoud.net 2021-2025, All rights reserved. 8
3 Description of the Hadith
Hadith (in Arabic: , transliteration: al-
prophet Muhammad (Pbuh) (Wiki2, 2012, p. 2) (Islahi, 1989). Hadith collections are regarded as
important tools for determining the Sunnah, or Muslim way of life, by all traditional schools of
jurisprudence. In Islamic terminology, the term hadith refers to reports about the statements or actions
of the Islamic prophet Muhammad, or about his tacit approval of something said or done in his presence
(Wiki2, 2012, p. 2) (Islahi, 1989). The text of the Hadith (matn) would most often come in the form of
a speech, injunction, proverb, aphorism or brief dialogue of the Prophet whose sense might apply to a
range of new contexts. The Hadith was recorded from the Prophet for a period of 23 years between 610
and 633 (after the birth of Christ).
Figure 3.1 A very old manuscript of al- hadith collection. Courtesy of the British Library.
© Copyright Sayoud.net 2021-2025, All rights reserved. 9
4 Who was the Author of the Quran?
Muslims believe that the holy Quran is from Allah (God) and that Muhammad was only the narrator
who recited the sentences of the Quran as written by Allah (God), but not the author. See what Allah
(God) says in the Quran book: « O Messenger (Muhammad)! transmit (the Message) which has been
sent down to you from your Lord. And if you do not, then you have not conveyed his Message. Allah
will protect you from people. Allah do not guide the people who disbelieve » [5:67].
Some doubts about the origins of the Holy Quran trying to find a human source for this book do exist.
Such assumptions suppose that the Holy Quran is an invention of the prophet Muhammad as reported
by Al-Shreef (Al-Shreef, 2009).
For a long time, different scientists tried to present strong context-based demonstrations showing that
this assumption is impossible.
The purpose of our research work is to conduct a text-mining based investigation in order to see
whether the two concerned books could statistically belong to the same author or not: i.e. authorship
discrimination (Mills, 2003) (Tambouratzis et al., 2000) (Tambouratzis et al., 2003), regardless of the
literal style or context.
© Copyright Sayoud.net 2021-2025, All rights reserved. 10
5 Stylometric Analysis and Automatic Authorship
Attribution
Individuals have distinctive ways of speaking and writing, as explained by Corney (Corney, 2003), and
there exists a long history of linguistic and stylistic investigation into authorship attribution (Holmes,
1998). In recent years, practical applications of authorship attribution have grown in areas such as
intelligence (linking intercepted messages to each other and to known terrorists), criminal law
(identifying writers of ransom notes and harassing letters), civil law (copyright and estate disputes),
and computer security (tracking authors of computer virus source code). As reported by Madigan
(Madigan et al., 2005), this activity is part of a broader growth within computer science of identification
technologies, including biometrics (retinal scanning, speaker recognition, etc.), cryptographic
signatures, intrusion detection systems, and others.
Stylometry or author recognition is a research field that consists in recognizing the authentic author of
a piece of text. It is evident that the recognition accuracy is not as high as some biometric modalities
that are used in security purposes, but it has been shown that for texts with more than 2500 tokens, the
recognition task becomes significantly accurate (Signoriello et al., 2005) (Eder, 2010).
Stylometry (or author recognition) can be divided into several research fields:
- Authorship Attribution (Sarwar et al., 2020), or identification, which consists in identifying the
author(s) of a text;
- Authorship verification (Kestemont et al., 2020), which consists in checking if a text claimed to
be written by somebody is really written by himself;
- Authorship discrimination (S. 2012] H. Sayoud, 2012), which consists in checking if two texts
are written by the same author or not;
- Authorship Indexing (Zangerle et al., 2020), which consists in segmenting a multi-author text
into several homogeneous segments and giving the identity of each author in those
homogeneous segments;
- Plagiarism detection (Muangprathub et al., 2021) (Zouhir et al., 2021), which consists in
checking if a piece of text has been picked from another author.
That is; determining the real author of a piece of text has raised several questions and problems for
centuries. Problem of authorship can be of interest not only to humanities researchers, but also to
politicians, historians and religious scholars in particular. Thorough investigative journalism, combined
with scientific analysis (e.g., chemical analysis) of documents has traditionally given good results
(Juola, 2006).
Furthermore, the recent development of improved statistical techniques in conjunction with the large
availability of digital corpora, have made the automatic and objective inference of authorship a practical
and easy task. That is why, this research field has seen an explosion of scholarship, resulting in several
related works (Farringdon, 1996) (Sari et al., 2018) (Foster, 2001) (Evert, 2017) (Love, 2002) (Al-
Batineh, 2019) (McMenamin, 2002) (Kalgutkar et al., 2018) (Mosteller & Wallace, 1964) (Schuster et
al., 2020).
Research works on authorship attribution usually appear at several types of debates ranging from
linguistics and literature through machine learning and computation, to law and forensics. Despite this
interest, the field itself is somewhat in confusion with a certain sense of best practices and techniques
(Juola, 2006).
Stylometry was also used in religious disputes. In fact, for the Christian religion, there exist several
disputes about the origin of some texts of the Bible. Such disputes are very difficult to solve due to the
delicacy of the problem, the religious sensitivity and because the texts were written a long time ago. For
example, early in the nineteenth century, Schleiermacher disputed the authorship of the Pauline Pastoral
Epistle 1 Timothy (Mills, 2003). As a result, other German speaking theologians, namely, F.C. Baur and
H.J. Holtzmann, initiated similar studies of New Testament books (Mills, 2003).
© Copyright Sayoud.net 2021-2025, All rights reserved. 11
In such problems, it is crucial to use rigorous scientific tools and it is important to interpret the results
very carefully.
Hence, knowing that authors possess specific stylistic features that make them differentiable (Li et al.,
2006), we tried to make some experiments of author discrimination between the Quran and some
results of these techniques confirm that supposition (Al-Shreef, 2009).
© Copyright Sayoud.net 2021-2025, All rights reserved. 12
6 The Quran and Hadith Corpora
Herein we summarize the size of the two books in terms of words, tokens, pages, etc.
6.1 Dimension of the two religious books
In a previous work, we used the entire text of the Quran (something like 315 A4 pages) but a small
collection of the Hadith (not exceeding 3 pages) only, due to the difficulty to find a book containing
s context the author was
strongly advised by some experienced stylometric researchers, who were working on Greek discourses,
to try to increase the size of the Hadith text, in order to get a consistent comparison between the two
investigated books. So, after a thorough investigation on the Hadith texts, the author managed to collect
a confident and consistent dataset, which is organized in a form that is more convenient (book gathering
pure Prophet statements, called Bukhari Hadith).
That is, the present section summarizes the size of the two new investigated books in terms of words,
tokens, pages, etc. The statistical characteristics of these two books are summarized as follows:
▪ Quran size in terms of tokens: 873
▪ Hadith size in terms of tokens (Bukhari Hadith): 23068
▪ Quran size in terms of different words = 13473
▪ Hadith size in terms of different words (Bukhari Hadith) = 6225
▪ Quran A4 pages in the Quran: 315 pages (subjective size)
▪ Hadith A4 pages in the Hadith (Bukhari Hadith): 87 pages (subjective size)
▪ Ratio of the Number of Quran Tokens / Number of Hadith Tokens = 3.79
▪ Ratio of the Number of Quran Lines / Number of Hadith Lines = 3.61
▪ Ratio of the Number of different Quran words / Number of different Hadith words = 2.16
▪ Ratio of the Number of Quran A4 Pages / Number of Hadith A4 Pages = 3.62
According to these size details, the two religious books seem relatively consistent, since the average
number of pages is 315 for the Quran book and 87 for the Hadith book. However, since the two books
do not have the same size, it will be necessary and prudent to segment these two books into segments
of more or less a same size, in order to avoid unbalanced results.
6.2 Segmentation
As quoted in section 6.1, the author already conducted an authorship investigation on the two religious
books by considering the whole books entirely (S. 2012] H. Sayoud, 2012). In that approach, when
comparing two books, it is difficult to know any part of the book is similar to the other one or different
from it. That is why 2 judicious segmentations have been proposed and applied on the different books,
© Copyright Sayoud.net 2021-2025, All rights reserved. 13
which consists in segmenting those books into several text segments, where the sizes of the segments
are more or less in the same range.
In the first segmentation technique, we divided every book into 4 different big text segments, where a
segment size is about 10 standard pages and all the segments are distinct and separated (without
intersection).
In the second segmentation technique, we used 14 different text segments for the Quran and 11 different
text segments for the Hadith, with approximately the same size (about 2100 words per dsegment). In
case of machine learning based classification, these segments should be organized as follows: some text
segments are selected from every book to represent the training data and the remaining text segments
are used during the testing step. However, in the Leave-One-Out method, all the text segments are used
for classification/attribution.
The segments have more or less the same size in terms of words as it is shown in table 6.1. The medium
size is about 2076 words per text.
The problem with such a size is that authorship attribution (AA) systems are usually not very accurate,
since it has been shown that the minimum text size, for a good AA process, is at least 2500 words per
size (Eder, 2010) (Signoriello et al., 2005).
Table 6.1: Sizes of the different text segments
Hadith text
segments
Size in
terms of
tokens
Quran
text
segments
Size in
terms of
tokens
H1
2035
Q1
2064
H2
2096
Q2
2071
H3
2053
Q3
2086
H4
2059
Q4
2085
H5
2081
Q5
2081
H6
2073
Q6
2080
H7
2031
Q7
2087
H8
2082
Q8
2074
H9
2088
Q9
2081
H10
2097
Q10
2079
H11
2083
Q11
2078
/
/
Q12
2092
/
/
Q13
2093
/
/
Q14
2081
6.3 Word structure of the different segments
A graphical representation of the word length frequency has been made for every text segment, in
order to see the overall structure of the used words in term of size. Figure 6.1 represents the smoothed
word length frequency curves versus the number of characters per word. It shows that the words have
more or less the same dimension frequency for both books, except for unigrams (1-character words),
© Copyright Sayoud.net 2021-2025, All rights reserved. 14
trigrams, tetragrams and octograms (8-character words), where we distinguish a clear difference in
their frequencies, but this observation cannot be used for objective discrimination purposes.
Figure 6.1: Word length frequency versus the word length (for all text segments). Q stands for a Quran
segment and H stands for a Hadith segment. Curves are obtained by interpolation.
1
6
11
16
21
26
1 3 5 7 9
H01 H02 H03 H04 H05 H06 H07
H08 H09 H10 H11 Q01 Q02 Q03
Q04 Q05 Q06 Q07 Q08 Q09 Q10
Q11 Q12 Q13 Q14
Frequency in %
Word length
© Copyright Sayoud.net 2021-2025, All rights reserved. 15
7 First Series of Experiments: Global Analysis
The first series of experiments analyses the two books in a global form (the text of every book is
analyzed as a unique big text). It concerns eight experiments. In this series of experiment we analyze
the two books (Quran and Hadith) in a global form.
7.1 Authorship Discrimination based on Words
This experiment represents an investigation on the word frequency. Results are displayed in the
following figure.
Notion of discriminative words
A particular interest concerns the discriminative words, as we can see in figure 7.1.
A discriminative word can be seen as a word that is frequently used in one text and rarely employed in
the other, which could represent a sample word that can be used for discriminating the two texts.
For instance, suppose that two authors are asked to write a letter in a same topic. Since every author has
a set of preferred words, one should retrieve some specific words that are commonly employed by one
author and almost never used by the other one.
Consequently one could distinguish the two authors (texts) by such discriminative words. That is why
that type of words is investigated in this section (see figure 7.1).
Figure 7.: Word frequency (in %) of some of the most discriminative words: black for the Quran and
grey for the Hadith.
For the words listed above, the frequencies are relatively different showing a dissimilarity between the
books vocabularies. Note that the first word in the left (namely
), And
very high frequency. It is used with a frequency of about 11% in the Quran and 8% in the Hadith, which
© Copyright Sayoud.net 2021-2025, All rights reserved. 16
involves an average relative difference of about 30%. The second word from the left (namely
),
Those
frequency (approximately 0%) in the Hadith, which represents an important discriminative word
between these books.
The first observation of these histograms shows that the two books are written by authors using different
vocabulary style.
7.2 Authorship Discrimination based on Characters
This experiment makes a comparison between the character frequencies of the two books. From the
resulting frequencies, we did a sorting of the differences between the 2 frequencies (Quran frequency
and Hadith frequency), for all the characters, in a descending order. At the end we kept only the 16 first
characters that have been sorted (see figure 7.2). In that figure, we have represented the character
frequencies used in the two books. We can see, for example, that for the first five characters (ie. , , ,
, and ), the difference between the utilization frequencies in the two books is appreciable. This
observation implies two different writing styles for the two books.
Figure 7.2 Frequency of the most discriminant Arabic characters
0
2
4
6
8
10
12
14
Quran Hadith
© Copyright Sayoud.net 2021-2025, All rights reserved. 17
7.3 Authorship Analysis based on COST parameter
Definition of the COST parameter
Usually, when poets write a series of poems, they make a termination similarity between the neighboring
sentences of the poem, such as a same final syllable or letter. To evaluate that termination similarity, a
new parameter estimating the degree of text chain (in a text of several sentences) has been proposed: the
COST parameter.
-
occurrence marks concern only the two last letters of the sentence.
It is interesting to note that Quran and Hadith books do not contain poems, but they consist in statements,
indications, histories, questions and answers, human obligations, advices, description of God,
description of the after-life, etc. The COST parameter, in this case, can give some information on the
structure of the text (ending structure). In this investigation, it has been employed to see if the two texts
respect certain regularities in the text structure or not and, if so, to assess the corresponding regularity
ratio.
For instance, let us observe the following English poem:
Never say it is the end when we do believe → COST = 2
And never accept that you do not retrieve → COST = 2
Life is so short to let things kill our mind → COST = 2
What to do in such situations dear friend → COST = 4
It is true that it is hard but victory will be in hand → COST = 2
Do not hesitate to try if you can make any change → COST = 1
Yes it is worth trying even if it is the last chance → COST = 1
So by counting the number of similar characters (i.e. (1+1) + (1+1) = 4), we get a COST value of 4. The
same procedure is repeated for each sentence until the last one.
© Copyright Sayoud.net 2021-2025, All rights reserved. 18
For concreteness, here are the COST values for some Hadith sentences (see table 7.3.1) and the COST
values of some Quran sentences (see table 7.3.2).
Table 7.3.1: COST values for some Hadith sentences
Sentence No
Cost
last 2 characters
Word
1\
0
2\
0
3\
0
4\
0
5\
0
6\
0
7\
0
8\
0
9\
0
10\
1
11\
1
12\
0
13\
1
14\
2
15\
1
16\
1
17\
1
18\
0
© Copyright Sayoud.net 2021-2025, All rights reserved. 19
Table 7.3.2: COST values of some Quran sentences
Sentence No
Cost
last 2 characters
Word
3116\
4
3117\
4
3118\
3
3119\
1
3120\
1
3121\
2
3122\
2
3123\
3
3124\
4
3125\
4
3126\
4
3127\
3
3128\
2
3129\
3
3130\
4
3131\
3
3132\
1
3133\
2
3134\
4
3135\
3
3136\
2
According to these tables, we remark that for the Hadith mixture, there are many COST values equal to
0; and when the COST is non-null, it has very small values: the average COST is only 0.46.
For the Quran, we notice that the COST is almost never null and the corresponding values are relatively
high: the average COST of the Quran is approximately 2.52.
This fact means that the structure of the Quran is very different from the Hadith one. Consequently, the
two books must have two different author styles.
© Copyright Sayoud.net 2021-2025, All rights reserved. 20
7.4 Word length frequency based analysis
This experiment represents an investigation on the word length frequency. Herein, we must define some
technical terms employed in our paper:
-The word length is the number of letters composing that word.
-
composed of n letters each, present in the text.
Figure 7.4: Word length frequency in histograms representation.
In figure 7.4, the two spectra are represented simultaneously, which gives an interesting way to compare
the two
observations related to every word length are given here below:
10.95
16.82
17.61
19.93
16.61
11.24
4.09
2.48
0.25
0.02
0
0
8.03
17.34
21.77
23.2
16.41
8.77
3.36
0.97
0.14
0.01
0
0
0
5
10
15
20
25
12345678910 11 12
Quran Hadith
© Copyright Sayoud.net 2021-2025, All rights reserved. 21
• Length 1: FQuran (1)=10.95%, whereas FHadith (1)=8.03%; which shows that the words composed
of a single letter are much more frequently used in the Quran than in the Hadith subset. For this
frequency we notice a great difference between the two books. The Pearson chi-square (uncorrected
for continuity) regarding this result is 167.54, involving a probability of consistency p < 0.0001,
consequently results related to 1-word frequency appear to be significant.
• Lengths 2, 3 and 4: For these cases, the Hadith subset contains many more words than the Quran.
We conclude that the Hadith subset uses much more short words than the Quran. The number of
short words in the Hadith subset is 62.31%, whereas, in the Quran, it is only 53.76%: namely a
difference of 8.55%. The Pearson chi-square (uncorrected for continuity) regarding this result is
468.37, involving a probability of consistency p < 0.00001, consequently results related to short-
word frequency appear to be significant.
• Lengths 5, 6, 7 and 8: For these cases, the Quran uses much more words than the Hadith subset. The
number of long words in the Quran is 34.42%, whereas, in the Hadith subset, it is only 29.51%:
namely a difference of 4.91%. The Pearson chi-square (uncorrected for continuity) regarding this
result is 198.3, involving a probability of consistency p < 0.0001, consequently results related to
long-word frequency appear to be significant.
Lengths 9 and 10: the Quran contains approximately a double number of words with 9 and 10 letters
than the Hadith. This fact shows that the Quran vocabulary contains more very-long words (very-long
stands for more than 8 letters) than the Hadith. The Pearson chi-square (uncorrected for continuity)
regarding this result is 10.78, involving a probability of consistency p < 0.001. Even though the
consistency probability is lower in this case, results related to very-long-word frequency appear to be
significant enough.
So, according to all these observations we conclude that the two authors have different styles.
7.5 Discriminative words based analysis
In this experiment, we look for the words that are present in one book and absent in the other.
Definition of "word": In our investigation, a word represents a sequence of characters linked to form
a noun, verb, complement, preposition, or a fusion of a preposition and another word (noun/verb) if they
are linked without space.
In this experiment, we analyze all the words present in the Hadith, and try to see if there is any
occurrence in the Quran. Similarly, on the other hand, we analyze all the words present in the Quran,
and try to see if there is any occurrence in the Hadith. If a word is present in only one book, it will be
retained; otherwise it will not be taken into consideration. The word can be a name, verb, complement
or a simple expression.
We recall that the part of the Bukhari Hadith contains 23068 tokens and 6225 different words. The
Quran contains 87339 tokens and 13473 different words.
Results of this experiment show that 62% of the Bukhari Hadith words are untraceable in the Quran and
83% of the Quran words are untraceable in the Bukhari Hadith (see figures 7.5.1 and 7.5.2). Such tokens
are called Discriminant Words (we chose this appellation due to the proposed application of
discrimination).
© Copyright Sayoud.net 2021-2025, All rights reserved. 22
Figure 7.5.1: Hadith words never used in Quran : 3885 different words (over 6225 total different words
contained in Bukhari Hadith) : 3885/6225 =62.41% of words absent in Quran.
Figure 7.5.2: Quran words never used in Bukhari Hadith: 11133 different words (over 13473 total
different words contained in Quran ) : 11133/13473=82.63% of words absent in Hadith.
Hadith words
absent in Quran
62%
Hadith words
present in Quran
38%
Hadith words
Quran words
absent in Hadith
83%
Quran words present in
Hadith
17%
Quran words
© Copyright Sayoud.net 2021-2025, All rights reserved. 23
Observation and Discussion
Practically, it is impossible for a same author to write two books (related to a similar topic) with a so
great difference in the vocabulary. Therefore, we can deduce that the two books should come from two
authors who are characterized by two different vocabularies.
7.6 Numbers citation based analysis
This experiment investigates the citation of numbers in the text: How many times a specific number (0..
9) has been used in the books?
Figure 7.6: Frequency of numbers citation with interpolated curves.
In the above graph (figure 7.6), we notice that the most frequently cited number, in the Quran, is the
We also notice that both books use more odd numbers than even ones, except for the Quran book
-null number, which is not the case
for the Hadith (its corresponding frequency is about 10%).
In this experiment, the difference regarding the use of numbers, between the two books, is so different
that we can state that the authors should probably be different.
0
5
10
15
20
25
30
35
40
45
012345678910
Quran Hadith
© Copyright Sayoud.net 2021-2025, All rights reserved. 24
7.7 Animal citation based analysis
The eighth experiment investigates the citation of animals in the text.
The animal citation frequency (freq) is defined as follows:
freq in % =100.(frequency of occurrence / total number of animal citations)
Example:
freq in Quran = 100.(frequency of occurrence / 155), since the total number of animal citations was
155.
freq in Hadith = 100.(frequency of occurrence / 94), since the total number of animal citations was 94.
First observation:
The following table 7.7.1 shows that for the seven following animals, the difference in citation between
the two books is relatively great:
• The name (General name of kamels, cows, sheeps)) is cited 33 times in the Quran,
whereas in the Bukhari Hadith it is cited only 2 times;
• The name (Dog) is cited only 5 times in the Quran, whereas in the Bukhari Hadith it is cited
13 times;
• The name (Sheep) is completely absent in the Quran, whereas in the Bukhari Hadith it is cited
10 times;
• The name (Animal ) is cited 17 times in the Quran, whereas in the Bukhari Hadith it is cited
only 3 times;
• The name (Camel ) is cited only 2 times in the Quran, whereas in the Bukhari Hadith it is
cited 7 times;
• The name (Calf ) is cited 10 times in the Quran, whereas in the Bukhari Hadith it is
completely absent;
• The name (Fish ) is cited only 4 times in the Quran, whereas in the Bukhari Hadith it is cited
8 times;
Table 7.7.1: Citation frequency of some animals appearing more frequently in one book than in
the other.
*****
Animal
Translation
Citation in
Quran
Citation in
Hadith
Frequency
in Quran
(%)
Frequency
in Hadith
(%)
General name
(kamels, cows,
sheeps)
21.3
2.13
Dog
3.2
13.83
Sheep
0.0
10.64
Animal
11.0
3.19
Camel
1.3
7.45
Calf
6.5
0
Fish
2.6
8.51
© Copyright Sayoud.net 2021-2025, All rights reserved. 25
Second observation:
In table 7.7.2, we quote the animals that are quoted in the Quran but completely absent in the Bukhari
Hadith. There are 29 such animal names.
We remark that several animal names are not cited in the Bukhari Hadith and particularly the name
(calf), which is cited 10 times in the Quran and which is completely absent in the Bukhari Hadith.
Table 7.7.2: Citation frequency of animals that are quoted in the Quran but completely absent in the
Bukhari Hadith.
**********
Third observation:
In table 7.7.3, we quote the animals that are quoted in the Bukhari Hadith but completely absent in the
Quran. There are 11 such animal names.
Animal
Translation
Citation in Quran
Citation in Hadith
Calf
Absent
Ant
Absent
Monkey
Absent
Female sheep
Absent
Snake
Absent
Fly
Absent
Spider
Absent
Grasshopper
Absent
Crow
Absent
Fast snake
2
Absent
Lion
Absent
hoopoe
Absent
snake
Absent
Bird
Absent
Type of horse
Absent
Mosquito
Absent
Bee
Absent
Lamb
Absent
Goat
Absent
Lice
Absent
Frog
Absent
Thirsty camel
1
Absent
General name
(kamels, cows,
sheeps)
Absent
Maybe: type of
birds
Absent
Lions
Absent
(
Earthworm
Absent
Pregnant camel
(+/-)
1
Absent
Wild animals
Absent
Type of wild
monkeys or
maybe zebras
1
Absent
© Copyright Sayoud.net 2021-2025, All rights reserved. 26
A particular observation can be done about the name (sheep), which is cited 10 times in the Bukhari
Hadith and which is completely absent in the Quran.
Table 7.7.3: Citation frequency of animals that are quoted in the Bukhari Hadith but completely absent
in the Quran.
Discussion: Results show that there are different animal name citations in the two books. That is, two
cases are possible:
- the two books could be related to two topics that are contextually different, citing a contextual
type of animal consequently;
- or the two authors should have different stylistic preferences for animal appellations and
citations.
However, when we read the two books, we notice that the topics are mainly the same. This fact proposes
that the second case is the most probable in this investigation.
7.8 Special Ending bigrams
This special investigation is made on six ending bigrams, which are often used in Arabic. The bigram
consists of a succession of two successive characters in the text.
character ---
--feed.
The different bigrams that have been chosen in this investigation are as follows:
""
""
""
" "
Animal
Translation
Citation in Quran
Citation in Hadith
Sheep
Absent
Bull
Absent
Cat
Absent
Bird
Absent
Lizard
Absent
Type of red
camels
Absent
1
Horse
Absent
Sheep
Absent
Rooster
Absent
Hen
Absent
Miraculous type
of horse (Buraq)
Absent
© Copyright Sayoud.net 2021-2025, All rights reserved. 27
""
""
Usually, these bigrams (except the 3rd and 4th ones) are often related to the plural form in Arabic.
Figure 7.8: Frequency of some ending bigrams.
We notice, in figure 7.8, that there is a great difference in the use of these ending bigrams between the
Quran (where the frequency is relatively high) and the Bukhari Hadith (where the frequency is relatively
low), especially for the two first bigrams and the two last bigrams.
This phenomenon can be justified by the fact that the Quran uses much more frequently the plural form
in its sentences.
So the authors of the two books appear to have different styles ofwriting: in the Quran, the plural form
is more employed than in Hadith.
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1
Bigram frequency %
Freq in Quran % Freq in Hadith %
© Copyright Sayoud.net 2021-2025, All rights reserved. 28
7.9 Discussion on the first series of experiments
This research work consists in an investigation of authorship discrimination between two old Arabic
books: the Quran and Bukhari Hadith.
In this first series of experiments, which consists 8 different experiments, we have analyzed the two
books in a global form. And amazingly, the 8 corresponding experiments have led to the same
conclusion: the two investigated books should have different authors.
Consequently, this first series of experiments shows that the Quran could not be written by the Prophet
Muhammad.
© Copyright Sayoud.net 2021-2025, All rights reserved. 29
8 Second series of experiments: Big Segments based
Segmental analysis
The second series of experiments analyses the two books in a segmental form with big text segments:
four different segments of texts are extracted from every book and the different texts are analysed and
compared.
In such tasks of authorship attribution or discrimination, several linguistic features have been proposed
by different researchers. We can quote four main types of these features:
Vocabulary based Features: A simple way to confirm or refute authorship is to look for something
that completely settles the authorship question (Juola, 2006). It is clear, then, that the individual words
an author uses can reveal his or her identity. The problem with such features is that the data can be
faked easily. A more reliable method would be able to take into account a large fraction of the words
in the document (Juola, 2006) as the average sentence length.
Syntax based Features: One reason that function words perform well is because they are topic-
independent (Juola, 2006)
One simple way to capture this is to tag the relevant documents for part of speech or other syntactic
constructions (Stamatatos et al., 2001) using a tagger.
Orthographic based features: One weakness of vocabulary-based approaches is that they do not take
(Juola, 2006).
Characters based features: Some researchers (Peng et al., 2003) have proposed to analyze documents
as sequences of characters. For example, the character 4-
example by all the words. That is why this type of parameter can replace several other high-level
linguistic features. Furthermore, several experiments showed that character n-gram is one of the most
reliable and robust features in authorship attribution (Stamatatos, 2009).
In this section, the author proposes some types of features and describes five related experiments: an
experiment using discriminative words, a word length frequency based analysis, an experiment using
the COST parameter, an investigation on discriminative characters and an experiment based on
vocabulary similarities.
In these experiments, the different segments are chosen as follows: one segment is extracted from the
beginning of the book, another one from the end and the two other segments are extracted from the
middle area of the book. A segment size is about 10 standard pages and all the segments are distinct
and separated (without intersection). These segments are denoted Q1 (or Quran 1), Q2 (or Quran 2),
Q3 (or Quran 3), Q4 (or Quran 4), H1 (or Hadith 1), H2 (or Hadith 2), H3 (or Hadith 3) and H4 (or
Hadith 4). Finally, these eight texts segments are more or less comparable in size.
8.1 Discriminative words
This first experiment investigates the use of some words that are very commonly used in only one of the
books. In practice, we remarked that the words: (in English: THOSE or WHO in a plural form)
© Copyright Sayoud.net 2021-2025, All rights reserved. 30
and (in English: EARTH) are very commonly used in the four Quran segments; whereas, in
the Hadith segments, these words are rarely used, as we can see in the following table (table 8.1).
Table 8.1: Some discriminative words and their frequencies.
Word
Frequency (%) in the Quran segments
Frequency (%) in the Hadith segments
Quran 1
Quran 2
Quran 3
Quran 4
Hadith 1
Hadith 2
Hadith 3
Hadith 4
1.35
1.02
1.12
0.75
0.11
0.03
0.02
0.08
0.34
0.63
0.59
0.42
0.23
0.13
0.18
0.15
For the frequency of occurrence is over 0.7% in the Quran segments, but it is between 0.02% and
0.11% in the Hadith segments (namely almost the 1/10th of the Quran frequency).
For the frequency of occurrence is about 0.5% in the Quran segments, but it is between 0.13%
and 0.23% in the Hadith segments (namely about the half).
These results show that the author of the Quran uses much more frequently these particular words than
the Hadith author does.
8.2 Word length frequency based analysis
The second experiment is an investigation on the word length frequency. In the following figure (figure
8.2), the different curves (smoothed curves), representing the « word length frequency » versus the
« word length », show the following two important points:
• The Hadith curves have more or less a gaussian shape that is pretty smooth; whereas the
Quran curves seem to be less Gaussian and present some oscillations (distortions).
• The Hadith curves are easily distinguishable from the Quran ones, particularly for the
lengths 1, 3, 4 and 8: for the lengths 1 and 8, Quran possesses higher frequencies, whereas
for the lengths 3 and 4, Hadith possesses higher frequencies.
The statistical consistency of the discrimination between the two groups, using frequency of
monograms, trigrams, tetragrams or octograms based words, which is evaluated with Fisher's exact test,
corresponds to a probability p of 2.86%.
© Copyright Sayoud.net 2021-2025, All rights reserved. 31
Figure 8.2: Word length frequency (smoothed lines).
Although these results cannot be used accurately in authorship discrimination, they can give preliminary
information on the sizes of the preferred words by each author. That is, according to these results we
should expect that the two text groups correspond to different authors.
8.3 COST parameter
he third experiment concerns the new COST parameter which appears non-null only in the Holy Quran,
as we can see in table 8.3. The COST parameter is explained in section 7.3.
In fact, it measures the termination similarity between the neighboring sentences of a text, such as a
same final syllable or letter. That is, the COST parameter gives us assessment on the text organization
in term of ending structure.
The following table shows the average COST values of the 8 different segments.
Table 8.3: Average COST values for the different segments.
0
5
10
15
20
25
0 1 2 3 4 5 6 7 8 9 10 11 12 13 14
H1
H2
H3
H4
Q1
Q2
Q3
Q4
Word length frequency
Word length
Quran 1
Quran 2
Quran 3
Quran 4
Hadith 1
Hadith 2
Hadith 3
Hadith 4
COSTaverage
2.2
2.6
2.6
2.38
0.46
0.47
0.43
0.47
© Copyright Sayoud.net 2021-2025, All rights reserved. 32
We notice that the average value of COST is practically constant for all the Quran segments: it is about
2.2 at the beginning of the Quran, 2.4 at the end and it is about 2.6 in the area of the middle.
Similarly, this parameter appears constant for all the Hadith segments: it is about 0.46.
In addition, we notice that the mean values of the COST for Quran and Hadith are very different. This
great difference involves distinctive writing styles for the two books (i.e. two different styles concerning
the sentence ending).
8.4 Discriminative characters
The fourth experiment investigates the use of some characters that are very commonly used in only one
of the books.
In reality, we limited our investigation to one of the most interesting character, which seems to be very
time (in English, it is equivalent to the consonant W when used as consonant; or the vowel U when used
as vowel).
Furthermore, this character is important because it also represents the preposition AND (in English),
which is widely used in Arabic.
So, by observing the table below, we notice that this character has a frequency of about 7% in all Quran
segments and a frequency of about 5% in all Hadith segments.
Table 8.4: Frequency of the character in the different segments.
Segment
Q1
Q2
Q3
Q4
H1
H2
H3
H4
Frequency of
character
7.73
7.11
6.91
7.04
5.19
5.45
4.72
5.33
This difference in the character frequency shows that the 2 authors do not employ the character in the
same proportion.
8.5 Vocabulary based similarity
The fifth experiment makes an estimation of the similarity between the vocabularies (words) of the two
books.
So, in this investigation we propose a new vocabulary similarity measure that we called VSM (ie.
Vocabulary Similarity Measure), which is defined as follows:
VSM (text1, text2) = [number of common words between the 2 texts] / [size(text1) . size(text2)]1/2
Typically, in case of 2 identical texts, this similarity measure will have a value of 1 (ie. 100%). Hence,
the higher this measure is, the more similar (in terms of vocabulary) the two texts are.
© Copyright Sayoud.net 2021-2025, All rights reserved. 33
We recall that there are four texts of the Quran and four texts of the Hadith that are more or less
comparable in size. The different inter-measures of similarity are represented in the following matrix
(similarity matrix), which is displayed in table 8.5.1.
Table 8.5.1: Similarity matrix representing the different VSM similarity measures between segments.
VSM in %
H1
H2
H3
H4
Q1
Q2
Q3
Q4
H1
100
32.89
31.43
28.22
20.93
19.86
19.38
19.86
H2
32.89
100
31.37
29.23
20.84
19.99
18.63
19.45
H3
31.43
31.37
100
29.17
19.77
19.88
18.90
18.96
H4
28.22
29.23
29.17
100
19.93
18.68
18.55
18.79
Q1
20.93
20.84
19.77
19.93
100
29.73
29.56
24.49
Q2
19.86
19.99
19.88
18.68
29.73
100
34.88
25.22
Q3
19.38
18.63
18.90
18.55
29.56
34.88
100
27.09
Q4
19.86
19.45
18.96
18.79
24.49
25.22
27.09
100
We notice that all the diagonal elements are equal to 100%. We do remark also that all the Q-Q
similarities and H-H similarities are high, relatively to Q-H or H-Q ones (Q stands for a Quran segment
and H stands for a Hadith segment). This means that the 4 segments of the Quran have a great similarity
in vocabulary and the 4 segments of the Hadith have a great similarity in vocabulary, too. On the other
hand it implies a low similarity between the vocabulary styles of the two different books. This deduction
can easily be made from the following simplified table, which represents the mean similarity measure
between one segment and all the segments of a given book.
Table 8.5.2 gives the mean similarity according to Quran or Hadith for each segment X (X=Qi or X=Hi
, i=1..4), which can be expressed as the average of all the similarities between segment X and the
different segments of a same book. This table is displayed in order to see if a segment is more similar to
the Quran family or to Hadith family.
Table 8.5.2: Mean VSM similarity in % between one segment and the different segments of a same
book.
Mean Similarity
with H segments
Mean Similarity
with Q segments
H1
30.85
20.01
H2
31.16
19.73
H3
30.66
19.38
H4
28.87
18.99
Q1
20.37
27.92
Q2
19.60
29.94
Q3
18.87
30.51
Q4
19.27
25.60
© Copyright Sayoud.net 2021-2025, All rights reserved. 34
Similarly, we remark that the intra-similarities (within a same book) are high: between 26% and 31%;
and that the inter-similarities (segments from different books) are relatively low: not exceeding 20%.
This observation shows that all the segments of a same book appear to have a unique origin and that the
two books should have two different author styles.
8.6 Discussion on the second series of experiments
This second series, consisting of 5 experiments, analyzed the two books in a segmental form by using
statistical techniques of stylometry, where the segments are quite big (four different big segments of
texts are extracted from every book).
Once again, the 5 corresponding experiments have led to the same conclusion: the authors of the
two investigated books are very different.
Consequently and once again, this second series of experiments shows that the Quran could not be
conceived by the Prophet Muhammad.
© Copyright Sayoud.net 2021-2025, All rights reserved. 35
9 Third series of experiments: Automatic authorship
attribution with several features and several
classifiers
The next series of experiments, which consists in an automatic authorship attribution (Sanderson &
Guener, 2006), analyses the two books in a segmental form by using several features (words, word n-
grams, characters, character n-grams and dis-legomena) (Clement & Sharp, 2003) and several classifiers
(Camberra distance, Cosine distance, RN cross entropy, Histogram distance, Intersection distance,
Kullback Leibler distance, Manhattan distance, KS distance, LDA analysis and Naive Bayes classifier,)
(Juola, 2009).
The sizes of the segments are more or less in the same range: four different text segments, with
approximately the same size, are extracted from every book (the same dataset as in experiment 8.5).
It concerns two experiments:
- In the first experiment, the first segment of each book is taken as reference. Hence there will be two
reference texts, one representing the Quran author and the other representing the Hadith author. The six
remaining texts (3 for each book) have to be classified into Quran class or Hadith class.
- The second experiment is similar to the first one except that the reference texts, here, are represented
by the second segments of the two books respectively.
In this series of experiments, the author employs the JGAA software (Juola, 2009) to make an automatic
classification of the eight texts by using different features and different classifiers.
Concerning the number of selected examples (an example refers to a word, character, etc.), we have
considered 2 cases: in the first case, we consider all the examples and in the second case, we keep only
the 50 most frequent ones.
Note: in the following paragraphs, a score of 100% means that all the Quran segments are classified as
Quran class and all the Hadith segments are classified as Hadith class, without any error of attribution.
© Copyright Sayoud.net 2021-2025, All rights reserved. 36
9.1 First experiment
In the first investigation, we consider the segments Q1 and H1 as reference texts for the Quran and
Hadith, respectively. Then, Q2, Q3, Q4, H2, H3 and H4 will be considered as unknown texts to be
classified according to Quran class or Hadith class. During the feature extraction step, two cases are
possible: employing all the features or employing the most frequent ones.
In this experiment, all the text segments have to be classified into two classes: Quran class or Hadith
class. Classification results (displayed in %) are reported in table 9.1.
Table 9.1: Precision of good classification of the different segments with several features and several
classifiers
Feature
Classifier
Charac
.
Bigram
Charac
-ter
Charac
Tetra-
gram
Charac
Tri-
gram
Dis
Lego-
mena
Word
Word
Bi-gram
Word
Tri-
gram
Word
Tetra-
gram
Number of
features
All
all
all
all
All
all
50 most
freq.
50 most
freq.
50 most
freq.
Camberra
distance
100%
50%
83%
100%
100%
100%
100%
100%
100%
Cosine distance
100%
100%
100%
100%
100%
100%
100%
100%
100%
RN cross
entropy
100%
83%
100%
100%
100%
100%
100%
100%
100%
Histogram
distance
83%
100%
100%
83%
100%
100%
100%
100%
100%
Intersection
distance
100%
50%
100%
100%
100%
100%
100%
100%
100%
Kullback Leibler
dist
83%
83%
100%
100%
100%
100%
100%
100%
100%
Manhattan
distance
100%
100%
100%
100%
100%
100%
100%
100%
100%
LDA analysis
83%
100%
100%
83%
100%
100%
100%
100%
100%
This experiment employing several features (words, word n-grams, characters, character n-grams and
dis-legomena) and several classifiers (Camberra distance, Cosine distance, RN cross entropy,
Histogram distance, Intersection distance, Kullback Leibler distance, Manhattan distance, KS distance,
LDA analysis and Naive Bayes classifier), shows clearly that the 4 Quran segments should belong to a
same author, the 4 Hadith segments should belong to the same author too and that these two authors are
likely to be different.
9.2 Second experiment
In the following investigation, we consider the segments Q2 and H2 as reference texts for the Quran and
Hadith, respectively. Then, Q1, Q3, Q4, H1, H3 and H4 will be considered as unknown texts to be
classified according to Quran class or Hadith class. As previously, during the features extraction step,
two cases are possible: employing all the features or employing the most frequent ones.
© Copyright Sayoud.net 2021-2025, All rights reserved. 37
Table 9.2: Precision of good classification of the different segments with several features and several
classifiers
Feature
Classifier
Charc
Bigra
m
Charc-
ter
Charc
Tetra-
gram
Chara
c. Tri-
gram
Dis
Lego-
mena
Word
Word
Bi-
gram
Word
Tri-
gram
Word
Tetra-
gram
Number of
features
All
all
all
all
All
all
50
most
freq.
50
most
freq.
50
most
freq.
Camberra
distance
100 %
100 %
100 %
100 %
100 %
100 %
100 %
100 %
100 %
Cosine
distance
100 %
100 %
100 %
100%
100 %
100 %
100 %
100 %
100 %
RN cross
entropy
100 %
100 %
100 %
100 %
100 %
100 %
100 %
100 %
83 %
Histogram
distance
100 %
100 %
100 %
100%
100 %
100 %
100 %
100 %
100 %
Intersection
distance
100 %
50 %
100 %
100%
100 %
100 %
100 %
100 %
100 %
Kullback
Leibler dist
100 %
100 %
100 %
100 %
100 %
100 %
100 %
100 %
83 %
Manhattan
distance
100 %
100 %
100 %
100%
100 %
100 %
100 %
100 %
100 %
LDA analysis
100 %
100 %
100 %
100 %
100 %
100 %
100 %
100 %
100 %
Also, in this experiment all the text segments have to be classified into two classes: Quran class or
Hadith class. Results of good classification, displayed in %, are reported in table 9.2.
As in the first investigation, this experiment employing several features (words, word n-grams,
characters, character n-grams and dis-legomena) and several classifiers (Camberra distance, Cosine
distance, RN cross entropy, Histogram distance, Intersection distance, Kullback Leibler distance,
Manhattan distance, KS distance, LDA analysis and Naive Bayes classifier), shows clearly that the 4
Quran segments should belong to the same author, the 4 Hadith segments should belong to the same
author too and that these two authors are very probably different.
Discussion on these two experiments (9.1 and 9.2): According to these two experiments, we can
clearly see that the classification accuracy for the two books is 100% with almost all features and all
classifiers. Consequently, we can statistically state that the two investigated books have two different
authors or at least two different styles.
9.3 Discussion on the third series of experiments
The third series of experiments could be considered as a continuation of the 2nd series of experiments
since it uses the same text segments (i.e. same segmented corpus).
It consists 2 experiments analyzing the two books in a segmental form by using nearest neighbor
techniques of classification and by employing several types of features.
This series of experiments is interesting since it shows a score of 100% in almost all the experimental
tests, which means that all Qi segments are similar one another by representing the Quran class, and all
Hi segments are similar one another by representing the Hadith class, while the Qi segments are
completely different from the Hi segments. This difference is probably due to a distinction between the
2 styles. Hence this result clearly confirms that the two books Quran and Hadith should have different
authors.
© Copyright Sayoud.net 2021-2025, All rights reserved. 38
10 Fourthseries of experiments: Short Segments based
Segmental Authorship Attribution
In this research part, the religious books are finely split and segmented into 25 different (relatively short)
medium documents.
In fact, 14 text segments are extracted from the Quran book and 11 text segments are extracted from the
Bukhari Hadith. These segments have more or less the same size in terms of words and the medium size
is about 2080 words per text segment. The Quran is
statements, we chose only the certified texts of the Bukhari book. That is, four series of experiments are
done and commented. The first experiment concerns several experiments of authorship attribution using
different state of the art features and classifiers, the second experiment analyses the different texts by
using a new parameter called COST, the third experiment consists in an authorship discrimination using
performs a hierarchical clustering on the 25 text segments, in order to assess the real number of clusters
(author styles) and to see if the hypothesis of a unique author is possible.
The text segments are organized as follows: three text segments are selected from every book to
represent the training data and the remaining text segments are used during the testing step. In the other
cases, all the text segments are used for classification/attribution. The segments have more or less the
same size in terms of words as it is shown in table 10.1. The medium size is about 2076 words per text.
The problem with such a size is that authorship attribution (AA) systems are usually not accurate, since
it has been shown that the minimum text size, for a good AA process, is at least 2500 words per size
(Eder, 2010) (Signoriello et al., 2005).
Table 10.1: Sizes of the different text segments
Hadith text segments
Size in terms of
tokens
Quran text
segments
Size in terms of
tokens
H1
2035
Q1
2064
H2
2096
Q2
2071
H3
2053
Q3
2086
H4
2059
Q4
2085
H5
2081
Q5
2081
H6
2073
Q6
2080
H7
2031
Q7
2087
H8
2082
Q8
2074
H9
2088
Q9
2081
H10
2097
Q10
2079
H11
2083
Q11
2078
/
/
Q12
2092
/
/
Q13
2093
/
/
Q14
2081
© Copyright Sayoud.net 2021-2025, All rights reserved. 39
Word structure of the different segments
A graphical representation of the word length frequency has been made for every text segment, in
order to see the overall structure of the used words in term of size. Figure 10.1 represents the
smoothed word length frequency curves versus the number of characters per word. It shows that the
words have more or less the same dimension frequency for both books, except for unigrams (1-
character words), trigrams, tetragrams and octograms (8-character words), where we often distinguish
a certain difference in their frequencies, but this observation cannot be used for objective
discrimination purposes.
Figure 10.1: Word length frequency versus the word length (for all text segments). Curves are
obtained by interpolation.
In the following sections, we will describe the four experiments that have been conducted on the two
religious books for a purpose of authorship discrimination.
10.1 Experiments of Authorship Attribution using a Hierarchical
clustering
In order to represent the stylistic similitude between the different texts, in a graphical way, a
hierarchical clustering (Sayoud 2012-b), using cityblock distance, has been performed on all text
segments by using the following features: COST parameter (see section 7.3) and frequency of the
displayed in figure 10.1, where we can see the different possible clusters and their costs (distances).
The smaller the cost is, the more similar the segments are (in the same cluster).
1
6
11
16
21
26
1 3 5 7 9
H01
H02
H03
H04
H05
H06
H07
H08
H09
H10
H11
Q01
Q02
Q03
Q04
Q05
Q06
Q07
Q08
Q09
Q10
Q11
Frequency in %
Word length
© Copyright Sayoud.net 2021-2025, All rights reserved. 40
Figure 10.2: Dendrogram of a hierarchical clustering corresponding to the 25 text segments.
As we can see in figure 10.2, the segments have been automatically divided into 2 main clusters:
gathering all the text segments of the Hadith. We can notice that the last clustering into one cluster
(big line at the top) is inconsistent for two reasons: first, because the corresponding distance of this last
cluster is more than 4.5, which is relatively very large; and second, because we do not retrieve any link
between heterogeneous segments at all (clusters grouping different label types such as Qj-Hk). This
result confirms the previous conclusions stating that the different text segments should belong to 2
different authors, or at least 2 different author styles. It also shows that Quran texts are relatively
similar (low intra-variability with distances less than 2) and that Hadith texts are relatively similar too
(low intra-variability with distances less than 1).
10.2 Experiments of Authorship Attribution using different state of
the art features and classifiers
This series of experiments, which consists in an authorship attribution (Sanderson & Guener, 2006),
analyses the two books in a segmental form by using several features (word n-grams, character n-grams
and rare words) (Clement & Sharp, 2003) and several classifiers: Stamatatos distance, Canberra
distance, Cosine distance, RN cross entropy distance, Intersection distance, Manhattan distance, SMO-
SVM (Sequential Minimal Optimization based Support Vector Machine) classifier, Linear Regression
classifier and MLP (Multi Layer Perceptron) classifier.
H09 H11 H08 H03 H07 H01 H05 H02 H10 H04 H06 Q01 Q06 Q03 Q04 Q07 Q09 Q10 Q13 Q14 Q11 Q12 Q02 Q05 Q08
0
0.5
1
1.5
2
2.5
3
3.5
4
4.5
Distance
Cluster Q
Cluster H
© Copyright Sayoud.net 2021-2025, All rights reserved. 41
10.2.1. Brief description of the different classifiers
Short definitions of the different classifiers are given below:
▪ Manhattan distance
This distance (S. 2012] H. Sayoud, 2012) is very reliable in text classification. The Manhattan distance
between two vectors f and g is given by the following formula:
(10.1)
where n is the length of the vector.
▪ Cosine distance
Cosine similarity is a measure of similarity between two vectors that measures the cosine of the angle
between them. The technique is also used to compare documents in text mining. The cosine of two
vectors can be derived by using the Euclidean dot product formula:
(10.2)
and magnitude as:
(10.3)
where denotes the magnitude of vector f and n is its length (Wiki_COS, 2013).
▪ Stamatatos distance
This distance was proposed by Stamatatos (Stamatatos, 2007). The Stamatatos distance between two
vectors f and g is given by the following formula:
(10.4)
where n is the length of the vector.
▪ Canberra distance
Canberra distance is a numerical measure of the distance between pairs of points in a vector space. It is
more or less similar to Manhattan distance. It is mostly used for data scattered around the origin. The
Canberra distance between two vectors f and g is given by the following formula:
© Copyright Sayoud.net 2021-2025, All rights reserved. 42
(10.5)
where n is the length of the vector.
▪ Cross Entropy Distance
The Cross entropy distance, where f and g are supposed independent (Juola, 2006), is given by:
(10.6)
It has been widely used (improved version) by Juola (Juola, 2006) in his released software.
▪ Intersection distance
The intersection distance, which measures the dissimilarity between two sample sets, is complementary
to the Jaccard coefficient and is obtained by subtracting the intersection-to-union ratio from 1:
(10.7)
▪ Multi-Layer Perceptron (MLP)
The MLP is a classic neural network classifier that uses the errors of the output to train the neural
network (H. Sayoud, 2003). The MLP can use different back-propagation schemes to ensure the training
of the classifier. The MLP is trained by the two first texts for every author, whereas the remaining text
(the third one) is used for the testing task. Usually the MLP is efficient in supervised classification,
however in case of local minima, we could get some errors of classification.
▪ The Sequential Minimal Optimization based Support Vector Machine (SMO-SVM)
In machine learning, support vector machines (SVMs) are supervised learning models with associated
learning algorithms that analyze data and recognize patterns, used for classification and regression
analysis. The basic SVM takes a set of input data and predicts, for each given input, which of two
possible classes forms the output, making it a non-probabilistic binary linear classifier. Given a set of
training examples, each marked as belonging to one of two categories, a SVM training algorithm builds
a model that assigns new examples into one category or the other. A SVM model is a representation of
the examples as points in space, mapped so that the examples of the separate categories are divided by
a clear gap that is as wide as possible. New examples are then mapped into that same space and predicted
to belong to a category based on which side of the gap they fall on. In addition to performing linear
classification, SVMs can efficiently perform non-linear classification using what is called the kernel
trick, implicitly mapping their inputs into high-dimensional feature spaces. The SVM is a very accurate
classifier that uses bad examples to form the boundaries of the different classes (Witten et al., 1999).
The SMO algorithm is used to speed up the training of the SVM (Keerthi et al., 2001).
▪ Linear Regression
The Linear Regression is the oldest and most widely used predictive model. The method of minimizing
the sum of the squared errors to fit a straight line to a set of data points was published by Legendre in
© Copyright Sayoud.net 2021-2025, All rights reserved. 43
1805 and by Gauss in 1809. Linear regression models are often fitted using the least squares approach,
(as with least absolute deviations regression), or by minimizing a penalized version of the least squares
loss function as in ridge regression (Wiki_REG, 2013) (Huang & Pan, 2003).
10.2.2. Authorship Attribution Results
As quoted previously, there are 25 different text segments of about 2080 words each, consisting of 11
Hadith segments and 14 Quran segments. In these experiments, 3 segments of the Hadith and 3 other
segments of the Quran are used for the training and the remaining segments (8 Hadith segments and 11
Quran segments) are used for the testing. Therefore, there are 19 different segments to identify according
to 2 referential Authors (Quran Author or Hadith Author).
Note: in the following paragraphs, an attribution error of 0% means that all the Quran segments are
classified as “Quran class” and all the Hadith segments are classified as “Hadith class”, without any
error of attribution. In fact the attribution error is defined as the ratio of the number of false attributions
over the total number of testing segments (see equation 10.8).
(10.8)
Table 10.1: Attribution error for the different text segments. There are 11 segments for the Hadith (8
testing + 3 reference) and 14 for the Quran (11 testing + 3 reference).
Feature
Classifier
Charac.
Bigram
Charac.
Tri-gram
Charac.
Tetra-gram
Word
Bi-gram
Word
Tri-gram
Word
Tetra-gram
Word
Rare words
(freq=1.. 3)
Number of
features
All
All
All
50 most
freq.
50 most
freq.
50 most
freq.
All
All
SMO-SVM
0%
0%
0%
0%
0%
0%
0%
0%
Linear Regression
0%
0%
0%
0%
0%
0%
0%
0%
MLP
0%*
0%*
0%*
0%
0%
0%
0%*
0%*
Stamatatos distance
0%
0%
0%
0%
0%
5.3%
0%
0%
Canberra distance
0%
0%
0%
0%
0%
10.5%
0%
0%
Cosine distance
0%
0%
0%
0%
0%
0%
0%
0%
RN cross entropy
0%
0%
0%
0%
5.3%
0%
0%
0%
Intersection distance
-
0%
0%
0%
0%
0%
5.3%
0%
Manhattan distance
0%
0%
0%
0%
0%
0%
0%
0%
* : means that only the 500 most frequent features are employed
- : means a classification failure
© Copyright Sayoud.net 2021-2025, All rights reserved. 44
By observing the above table (table 10.1), we can notice that all Quran segments are attributed to the
Once again, from this result, we can deduce that the 2 religious
books should have 2 different authors (or at least 2 different writing styles) and that every book should
be written by one author (or at least one writing style).
10.3 Experiments of Authorship Attribution using the COST parameter
The COST parameter has already been explained in section 7.3 with several examples. This parameter
measures the degree of similarity between sentences endings.
According to the previous results of section 7.3, we could remark that for the Hadith text, there were
many COST values equal to 0; and when the COST is non-null, it has very small values: the average
COST is only 0.46. For the Quran, we noticed that the COST is almost never null and the
corresponding values are relatively high: the average COST of the Quran is approximately 2.52. This
interesting fact suggests the application of this type of experiment on the different text segments in
order to see if there exists a stylistic difference between those segments. The resulting average COST
values are represented in figure 10.3.
Figure 10.3: Average COST values for all text segments
0
0.5
1
1.5
2
2.5
3
H01
H02
H03
H04
H05
H06
H07
H08
H09
H10
H11
Q01
Q02
Q03
Q04
Q05
Q06
Q07
Q08
Q09
Q10
Q11
Q12
Q13
Q14
COST
© Copyright Sayoud.net 2021-2025, All rights reserved. 45
Figure 10.3 shows a sharp difference between the Quran segments, which present relatively high COST
values, and the Hadith segments, for which the COST values are very small. This fact implies that the
structures of Quran and Hadith are different. Consequently, and since we deal with the same topic (i.e.
religion), the two books should have two different author styles.
Furthermore, in order to assess the significance of the previous results, a statistical investigation on the
consistency of the discrimination between the two types of segments, is made by using the Fisher's
statistical exact test (Lowry 2012). Results show a two-tailed P probability that is less than 0.0001. This
result shows that the association between the style and COST parameter is statistically significant.
10.4 Experiments of Authorship Attribution using the frequency of
the word “”
This experiment investigates the use of some words that are very commonly used in only one of the
books (H. Sayoud, 2012). In practice, we remarked that the word (in English: THOSE or WHO in
a plural form) is very commonly used in the Quran; whereas, in the Hadith, this word is rarely used, as
we can see in the following figure. Its occurrence frequency is between 0.63% and 2.02% for Quran
segments, but it is between 0% and 0.29% for Hadith segments (see figure 10.4). Its average occurrence
frequency is 1.3% for Quran segments and it is only 0.09% for Hadith segments (namely almost the
1/14th of the average Quran frequency).
These results show that the author of the Quran uses much more frequently this particular word than the
Hadith author does.
Figure 10.
0
5
10
15
20
25
30
35
40
45
H1
H 2
H3
H4
H5
H6
H7
H8
H9
H10
H11
Q1
Q2
Q3
Q4
Q5
Q6
Q7
Q8
Q9
Q10
Q11
Q12
Q13
Q14
Numbrer of citations
© Copyright Sayoud.net 2021-2025, All rights reserved. 46
As previously, in order to evaluate the statistical significance of these results, a Fisher's statistical
exact test (Lowry 2012) has been made to compute the discrimination consistency. We get a two-tailed
P probability that is less than 0.0001. This result means that the association between style and citation
10.5 Discussion on the fourth series of experiment
As a continuation of the previous experiments on the authorship analysis of the holy Quran, in the
present investigation we have performed a segmental analysis for the task of authorship discrimination
(Tambouratzis et al., 2000) (Tambouratzis et al., 2003) between the two old Arabic religious books.
That is, four series of experiments have been made:
The first series of experiments performs a hierarchical clustering on the 25 text segments, in order to
see how many possible clusters really exist and if the hypothesis of a unique author is possible.
The second series of experiments consists in an authorship attribution task, which analyses the different
text segments by using several state-of-the-art features and classifiers.
he third series of experiments concerns the new COST parameter, which appears non-null only in the
Quran. This parameter estimates the degree of similarity between the endings of sentences.
The fourth series of experiments investigates the use of some words that are very commonly employed
in only one of the books. In this research work, the word: (in English, it is translated into: THOSE
or WHO in its plural form) has been chosen and analysed in the different text segments.
After observing all the experimental results and since the two books appear to have the same theme (i.e.
religion), it would be reasonable to deduce the following conclusions:
• The two segmented books should have different authors (or at least two different author
styles);
• All the segments that have been extracted from a unique book (from the Quran only, or from
the Hadith only) should probably belong to the same author.
According to these two important results, we should be able to extend our conclusions to the entire
books from which the concerned segments were extracted. In fact the styles of these text segments
represent the style of their corresponding original books (i.e. statistically). Consequently, it appears that
the two investigated books should have different authors. Without entering in theological debates, the
present investigation gives us a new scientific way to analyze and check the authorship authenticity of
old or disputed documents.
© Copyright Sayoud.net 2021-2025, All rights reserved. 47
11 Fifth Series of Experiment: Stylometric Comparison
between the Quran and Hadith based on Successive
Function Words
Usually, when one writes a text document, several function words are put within that text to make a logical
link between words and to give more explanation to the major idea of the paragraph.
However, it is not very common to see two Successive Function Words (SFW) put together in a sentence,
-af
In this work, those successive function words are investigated in the Quran, Hadith and five other religious
books, in order to see whether the corresponding styles are similar (i.e. Stylometric comparison between the
different books).
The first results of this investigation show that the use of SFW is very discriminative between the Quran
and Hadith.
11.1 Used function words
Function words are very used in Arabic (El-Zohairy, 2008) (García-Barrero et al., 2012) (Almujaiw, 2017),
however successive pairs of function words cannot be translated directly to English, because of the complex
translation process and the different morphologies in Arabic and in English. So, for instance, the succession
of the following function words: (rough translation: From After) cannot be translated into English by
keeping both prepositions, which may lead to an ambiguous meaning (i.e. From after), while the correct and
most suitable translation is simply: After.
That is, 10 different types of Successive Function Words (SFW) have been selected from the Quran, Hadith
and five other religious books by computing their corresponding frequencies, in a purpose of authorship
discrimination (Sayoud, 2012) (Tambouratzis et al., 2004). Those successive words are represented in table
11.1 and their corresponding frequencies are given in table 11.2 and 11.3.
Table 11.1: The used Arabic successive words (SFW)
prep +prep
Translation
And Then-When-Since
And From-what
And Already
And-do-you-think-that Who
After / From After
That (is) for (whom/who)
That (is) because
And If
What (is) Not
Before / From Before
© Copyright Sayoud.net 2021-2025, All rights reserved. 48
11.2 Analysis of the books based on the SFW
11.2.1. Experiment 1 - Comparison between the Quran and Hadith
The main goal of this investigation is to see whether the two religious books Quran (Nasr, 2004)
(Ibrahim., 1997) and Hadith (Islahi, 1989) come from the same author or not, which could shed some
light on the authorship origin of the holy Quran.
The occurrence frequencies of the quoted pairs of function words, between the Quran and Hadith, are
given in table 11.2, where one can see that for the Quran the frequencies values are relatively high, while
for the Hadith, the frequencies are very low and sometimes null. This difference can be better noticed
in figure 11.1.
Table 11.2: Frequencies of the Arabic successive words (SFW) in the Quran and Hadith.
SFW
Frequency in
the Quran x10-4
Frequency in
the Hadith x10-4
8.24
0.00
1.26
0.00
14.77
0.43
0.23
0.00
15.23
0.87
0.69
0.00
1.03
0.00
3.43
0.43
4.35
0.43
10.65
0.00
Mean value
5.99
0.22
Figure 11.1: Graphical representation of the frequencies of the Arabic successive words (SFW)
We notice that the use of SFW is very different between the Quran and Hadith, showing the simple and
easy style of the Hadith while for the Quran the style appears to be more sophisticated and richer by
using much more successive function words.
0
2
4
6
8
10
12
14
16
Frequency x 104
Hadith Quran
© Copyright Sayoud.net 2021-2025, All rights reserved. 49
11.2.2. Experiment 2 - Comparison of seven different books: Quran, Hadith and five
other books
In this experiment, we try to compare the SFW frequencies of 7 different book belonging to different
authors. The occurrence frequencies of the quoted pairs of function words, between the different books:
Quran, Hadith and 5 other mixed Arabic books written by 5 religious scholars (i.e. Abdalkafy, Alghazali,
Alqaradawi, Alqarn and Amrokhaled) (Hadjadj & Sayoud, 2018), are given in table 11.3. One can see
that for the Quran, the frequencies values are relatively high, while for all other books, the frequencies
are very low and sometimes null. This difference is easily noticeable in figure 11.2.
Table 11.3: Frequencies of the Arabic successive words (SFW) in the seven books.
SFW
Frequency in
the Quran x10-4
Frequency in
the Hadith x10-4
Medium
Frequency in
the 5 other
books x10-4
8.24
0.00
0.00
1.26
0.00
0.59
14.77
0.43
1.32
0.23
0.00
0.00
15.23
0.87
0.92
0.69
0.00
0.00
1.03
0.00
0.00
3.43
0.43
0.00
4.35
0.43
0.92
10.65
0.00
1.76
Mean value
5.99
0.22
0.55
Figure 11.2: Graphical representation of the frequencies of the Arabic successive words (SFW)
As an overall consequence, one can conclude that the Quran style uses much more frequently the
combination of function words than all other books (i.e. the Hadith and 5 religious books). In other
0
5
10
15
20
Frequency x 104
Hadith
5 other books
Quran
© Copyright Sayoud.net 2021-2025, All rights reserved. 50
words, the SFW frequency is much higher in the holy Quran that in all other books that have been
analyzed in this investigation, which involves a particular and distinct style for the holy Quran.
11.3 Discussion on this investigation
By observing figure 11.1, one can notice that the utilization of SFW in the Hadith is very rare: the mean
frequency is only 0.22 10-4, while in the case of the holy Quran, the frequency is much higher: mean
frequency of about 6 10-4, which is 27 times higher than the previous one.
Similarly, by observing figure 11.2, one can notice that the utilization of SFW in the 6 religious books
(i.e. the Hadith and 5 other books) is very rare too. In fact, the mean frequency does not exceed 0.55 10-
4 for those books, while in the case of the holy Quran, the frequency is much higher: mean frequency
of about 6 10-4, which is 11 times higher than the mean frequency of the 5 religious books and 27 times
higher than the frequency of the Hadith.
This noticeable difference in the use of SFW has two important interpretations:
• Firstly, the two writing styles of the Quran and Hadith appear different, with regards to the use
of SFW.
• Secondly, the writing style of the Quran is even different from all the human religious books
that have been investigated, with regards to the SFW.
different from the Prophet style, but also different from all the Arabic styles that have been studied in
this investigation. This result suggests that probably the Quran Author style is beyond the traditional
human religious styles that are usually employed in the religious Arabic literature (ancient or
contemporary).
© Copyright Sayoud.net 2021-2025, All rights reserved. 51
12 Sixth Series of Experiment: Authorship Identification
of 7 Books – A Fusion Approach
In this investigation, we conduct some experiments of automatic authorship attribution on seven Arabic
religious books, namely: the holy Quran, Hadith and five other books written by five religious scholars.
The Arabic styles are almost the same (i.e. Standard Arabic) for the seven books. The genre is quite the
same and the topics of the different books are also the same (i.e. Religion).
The authorship characterization is based on four different features: character trigrams, character
tetragrams, word unigrams and word bigrams. The task of authorship identification is ensured by four
conventional classifiers: Manhattan distance, Multi-Layer Perceptron, Support Vector Machines and
Linear Regression. Furthermore, a fusion approach has been proposed to enhance the performances of
authorship attribution, with two fusion techniques.
A particular application is dedicated to the authorship discrimination between the Quran and Hadith, in
order to see if the two books could have the same author or not.
12.1 Corpus of the seven religious books
As cited previously, there are seven different books written by seven different authors: the holy Quran,
Hadith and 5 other books written by 5 religious scholars. We recall that the Arabic styles are almost the
same (i.e. Standard Arabic) for the 7 books, the genre of the books is the same and the topics are also
the same (i.e. Religion). We called this dataset: SAB-1 (Seven Arabic Books dataset One). These
books are described as follows:
1st book: the holy Quran (author: God (Allah)). The Quran is considered to be written by Allah (God)
and only sent down to the Prophet Muhammad fourteen centuries ago.
2nd book: the Hadith (author: the Prophet Muhammad) contains the statements of the Prophet
Muhammad. As previously, in this investigation, we used the Bukhari Hadith.
The 5 other books: represent books and texts collections written by 5 religious scholar, namely:
Mohammed al-Ghazali al-Saqqa (al-Ghazali, 2021), Yusuf al-Qaradawi (Alqaradawi, 2013), Omar
Abdelkafy (Abdelkafy, 2013), Aaidh ibn Abdullah al-Qarni (Al-Qarni, 2021) and Amr Mohamed
Helmi Khaled (Amr-Khaled, 2013).
Those seven books are preprocessed and segmented into different and distinct text segments. Every
segment is about 2900 tokens each. Here are the numbers of segments by book:
Table 12.1: Books specifications of SAB-1 dataset.
Book/Author
Number of segments
by book*
Big/ Small
parameter#
Training
set size
Testing set
size
1st book: the holy Quran
30 segments
Big
7
23
2nd book: the Hadith
8 segments
Small
4
4
3rd book: books of Alghazali
39 segments
Big
7
32
4th book: books of AlQuaradhawi
13 segments
Small
4
9
5th book: books of Abdelkafy
10 segments
Small
4
6
6th book: books of Aid Alkarny
23 segments
Big
7
16
7th book: books of Amrokhaled
9 segments
Small
4
5
*Each segment is composed of 2900 tokens.
#Big/Small is a logical parameter (i.e. binary value).
© Copyright Sayoud.net 2021-2025, All rights reserved. 52
The corpus is decomposed into 2 parts: training part and testing part, and since the different books have
different sizes, an optimal logical rule has been established: 4 text segments are used for the training of
small books and 7 text segments are employed for the training of big books. The main reasons for this
choice are explained here below.
The choice of the training dataset size is defined by a particular logical (binary) parameter we called
Big/Small, which gives a qualitative estimation on the size of the book. That is, if the size of the book
is over 20 segments, then it is considered as a big dataset otherwise it is considered small. The value or
the threshold 20 is equal to the half size of the biggest dataset (ie. 39 segments for Alghazali book, which
implies a threshold of 39/2 20). This scheme permits us to have different possible sizes for the training
dataset.
books. In fact, the value 4 is equal to the half size (50%) of the smallest book (ie. the smallest book
contains only 8 segments).
books. In fact, the value 7 is equal to the maximum size of the training set for the smallest book ( ie. a
maximum of 7 segments for the training, since we require at least 1 segment for the testing ).
These two training rules could be applied to the different books with regards to the parameter Big/ Small.
But even though, the value 7 is a limit that we cannot exceed (and could be seen as a fixed choice), we
cannot say that the value 4 is optimal for small texts: why not 3 or 5 text segments, for instance.
In order to check if this choice was judicious or not, (experimentally speaking), we did some experiments
of authorship attribution on another corpus consisting of 7 different books (from a second different
dataset called SAB- where the sizes of the books are very similar to
those of the previous one: SAB-1 dataset (see table 12.2). The used classification technique was based
on the Manhatan Centroid distance.
Table 12.2: Features of the second dataset: SAB-2*
Case 1
Case 2
Case 3
Book designation
Big/ Small
dataset
Training set size
Training set size
Training set size
Book A
Big
7
7
7
book B
Small
3
4
5
book C
Big
7
7
7
book D
Small
3
4
5
book E
Small
3
4
5
book F
Big
7
7
7
book G
Small
3
4
5
* Note that the corpus SAB-2 will no longer be utilized in the next sections.
Hence, three cases are investigated:
- Case 1: 3 text segments are used for the training of small books and 7 text segments are
employed for the training of big books;
- Case 2: 4 text segments are used for the training of small books and 7 text segments are
employed for the training of big books;
- Case 3: 5 text segments are used for the training of small books and 7 text segments are
employed for the training of big books.
The different results of authorship attribution, got on this second dataset, are summarized in the
following table:
© Copyright Sayoud.net 2021-2025, All rights reserved. 53
Table 12.3: Results of authorship attribution got on the second dataset: SAB-2*
Score of good attribution in % (experiments conducted on another corpus)
Training size
Char.
Bi-
gram
Char.
Tri-
gram
Char.
Tetra-
gram
Word
Word
Bi-
gram
Word
tri-
gram
Word
Tetra-
gram
Average
performance
in %
Case
Big
Small
Case 1
7
3
74.74
83.83
89.89
94.94
94.94
32.32
33.33
63.88
Case 2
7
4
76.84
89.47
91.57
93.68
97.89
54.73
31.57
69.47
Case 3
7
5
76.92
85.71
89.01
95.6
97.8
35.16
32.96
65.38
* Note that the corpus SAB-2 will no longer be utilized in the next sections.
According to this table (table 12.3), the case 2, corresponding to 4 training texts for the small books,
seems to be the most interesting case. That is, by observing the average performance given by Manhattan
distance, we can easily see that the best average score is 69.47%, which corresponds to the second case
(ie. 4 text segments for small books and 7 ones for big books). According to this result, the chosen
training configuration seems to be judicious and interesting for the authorship attribution experiments
conducted on the first dataset.
However, we should note that we cannot expand this result to other classifiers like machine learning
ones, especially those which need a great amount of training data, such as neural networks or support
vector machines, for instance.
12.2 Authorship Attribution Methods
Several experiments of authorship attribution are conducted on the 7 segmented religious books.
For a purpose of feature selection and evaluation (Hawashin et al., 2013), four types of characteristics
are employed: character-trigram, character tetra-gram, word and word-bigram. Two of these features
are based on characters and the two others are typically lexical.
Also, four different classifiers are used for the automatic authorship classification (into ideally 7
different classes), where every class should represent one particular author. The different classifiers are
defined as follows:
- Manhattan centroid distance;
- Multi Layer Perceptron;
- SMO based Support Vector Machines;
- Linear Regression.
Furthermore, a Fusion approach is proposed to try enhancing the attribution accuracy of the conventional
classifiers/features.
12.2.1 Conventional Classifiers
The 4 conventional classifiers are described here below.
▪ Manhattan distance
This distance (H. Sayoud, 2012) is very reliable in text classification. The corresponding distance
between two vectors X and Y is given by the following formula:
© Copyright Sayoud.net 2021-2025, All rights reserved. 54
(12.1)
where n is the length of the vector.
In this investigation, the different samples of the training are employed to build the centroid vector,
which will be used, as reference, to compute the required distance with the previous formula (also called
KNN method). Manhattan distance is simple to implement and very efficient for text classification.
▪ Multi-Layer Perceptron (MLP)
The MLP (Multi-Layer Perceptron) is a classical neural network classifier that uses the errors of the
output to train the neural network (H. Sayoud, 2003). The MLP can use different back-propagation
schemes to ensure the training of the classifier. It is trained by the different texts of the training set,
whereas the remaining texts are used for the testing task. Usually the MLP is efficient in supervised
classification, however in case of local minima; we usually can get some errors of classification.
▪ Sequential Minimal Optimization based Support Vector Machine (SMO-SVM)
In machine learning, support vector machines (SVMs) are supervised learning models with associated
learning algorithms that analyze data and recognize patterns, which are used for classification and
regression analysis. The basic SVM takes a set of input data and predicts, for each given input, which
of two possible classes forms the output, making it a non-probabilistic binary linear classifier. Given a
set of training examples, each marked as belonging to one of two categories, a SVM training algorithm
builds a model that assigns new examples into one category or the other. A SVM model is a
representation of the examples as points in space, mapped so that the examples of the separate categories
are divided by a clear gap that is as wide as possible. New examples are then mapped into that same
space and predicted to belong to a category based on which side of the gap they fall on.
In addition to performing linear classification, SVMs can efficiently perform non-linear classification
using what is called the kernel trick, implicitly mapping their inputs into high-dimensional feature
spaces.
The SVM is a very accurate classifier that uses bad examples to form the boundaries of the different
classes (Witten et al., 1999). Concerning the Sequential Minimal Optimization (SMO) algorithm, it is
used to speed up the training of the SVM (Keerthi et al., 2001).
▪ Linear Regression
Linear Regression is the oldest and most widely used predictive model. The method of minimizing the
sum of the squared errors to fit a straight line to a set of data points was published by Legendre in 1805
and by Gauss in 1809. Linear regression models are often fitted using the least squares approach, but
with least absolute deviations regression), or by minimizing a penalized version of the least squares loss
function as in ridge regression (Wiki_REG, 2013) (Huang & Pan, 2003).
12.2.2 The Fusion approach
In order to enhance the authorship attribution performance, we have proposed the use of several
classifiers, which are combined in order to get a lower identification error: this combination is
technically called Fusion (Tchechmedjiev et al., 2013).
© Copyright Sayoud.net 2021-2025, All rights reserved. 55
Theoretically, the fusion can be performed at different hierarchical levels and forms. A very commonly
encountered taxonomy of data fusion is given by the following techniques (Jain et al., 2004) (Dasarathy,
1994) (Verlinde, 1999):
• Feature level where the feature sets of different modalities are combined. Fusion at this level
provides the highest flexibility but classification problems may arise due to the large
dimension of the combined (concatenated) feature vectors.
• Score (matching) level is the most common level where the fusion takes place. The scores
of the classifiers are usually normalized and then they are combined in a consistent manner.
• Decision level where the outputs of the classifiers establish the decision via techniques such
as majority voting. Fusion at the decision level is considered to be rigid for information
integration (Stylianou & et al., 2005), but it is not complicated in implementation.
In this investigation, we propose the use of the third technique, namely the decision level based fusion.
Furthermore, two types of combinations are employed: combination of features, called FDF or Feature-
based Decision Fusion, and combination of classifiers, called CDF or Classifier-based Decision Fusion.
- Feature-based Decision Fusion (FDF): In the first proposed fusion (combination of several features),
three different features are employed:
-Character-tetragram;
-Word;
-Word Bigram.
The fusion technique fuses the different corresponding scores of decision into one decision (the final
decision). The chosen classifier is Manhatan centroid because it has shown excellent performances
during the previous experiments.
It is called Feature-based Decision Fusion or FDF (see figure 12.1) and consists in fusing the outputs of
the different classifiers according to a specific vote provided by their different decisions: each decision
concerns one feature Fj.
The fused decision Df of N features is given by the following equation:
Decision = Df , with (12.2)
freq denotes the occurrence frequency of a specific decision and j=1..N.
Figure 12.1: Principle of the Feature-based Decision Fusion (FDF)
PCA reduction
PCA reduction
Feature FN
Clasifier 1 O1
Clasifier 2 O2
Clasifier Xo DN
Feature F1
Feature F2
Clasifier Xo D1
Clasifier Xo D2
Decision = Df
with
Authorship
Attribution
Decision
© Copyright Sayoud.net 2021-2025, All rights reserved. 56
- Classifier-based Decision Fusion (CDF): In the second proposed fusion (combination of several
classifiers), three different classifiers are employed:
-Manhattan centroid;
-SMO-SVM;
-MLP.
As previously, the fusion technique fuses the different corresponding scores of decision into one
decision (the final decision). Concerning the choice of the features, the word descriptor has been used
because it has been shown that this type of feature presented relatively good performances during our
experiments.
It is called Classifier-based Decision Fusion or CDF (see figure 12.2) and consists in fusing the outputs
of the different classifiers according to a specific vote provided by their different decisions: each
decision concerns one classifier Cj.
The fused decision Df of M classifiers is given by the following equation:
Decision = Df, with (12.3)
freq denotes the occurrence frequency of a specific decision and i=1..M.
Figure 12.2: Principle of the Classifier-based Decision Fusion (CDF)
All the results of the fusion approach are represented in tables 12.12 and 12.13, summarizing the
corresponding AA scores of the first and second fusion techniques respectively.
12.3 Experiments of authorship attribution
As mentioned previously, seven Arabic religious books are investigated and analyzed in order to make
a classification of the text documents per author: the experimented corpus is called SAB-1. We also
recall that several features and several classifiers are used in the experiments of authorship attribution.
PCA reduction
PCA reduction
Feature F0
Clasifier 1 O1
Clasifier 2 O2
Clasifier XM DM
Feature F0
Feature F0
Clasifier X1 D1
Clasifier X2 D2
Decision = Df
with
Authorship
Attribution
Decision
© Copyright Sayoud.net 2021-2025, All rights reserved. 57
12.3.1 Experiments of authorship attribution using conventional features and classifiers
In this section we report the different results obtained by using conventional classifiers and features.
The different experimental results are organized into 4 tables (table 12.8, 12.9,12.10 and 12.11):
- Table 12.8 displays the different results obtained with the Character-trigram feature;
- Table 12.9 displays the different results obtained with the Character-tetragram feature;
- Table 12.10 displays the different results obtained with the Word (Word-unigram) feature;
- Table 12.11 displays the different results obtained with the Word-bigram feature.
Those tables display the errors of authorship attribution given by the 4 classifiers: Manhatan centroid,
MLP, SMO-
summarizes the overall error of attribution for the 7 books. This indication gives us an interesting idea
on the overall performances of authorship attribution (corresponding to a specific feature).
Table 12.8: Identification Error in % using the feature: Character-trigram, on SAB-1 dataset.
Total
Identifica-
tion error on
the 7 books
The holy
Quran
book
The
Hadith
book
Aaid’s
book
Abd-
elkafy’s
book
Algha-
zali’s
book
Alquara
-dawi’s
book
Amro-
Khaled’s
book
Date / Century
Ancient :
6th
century
Ancient:
6th
century
Recent:
20th
century
Recent:
20th
century
Recent
:
20th
centur
y
Recent:
20th
century
Recent:
20th
century
Classifier
Manhatan
centroid
distance
4.2%
0%
0%
12.5%
0%
0%
22.2%
0%
MLP
classifier
3.1%
0%
0%
0%
16.7%
0%
22.2%
0%
SMO-SVM
classifier
4.2%
0%
0%
0%
33.3%
0%
22.2%
0%
Linear
Regression
4.2%
0%
0%
6.25%
16.7%
0%
22.2%
0%
In table 12.8, we can see that the best classifier is the MLP, which gives an error of only 3.1% (look at
the 1st columbn), the other classifiers have the same performances (total identification errors of 4.2%).
The two authors: Abdelkafy and Alquaradawi present some problems of authorship attribution, with
respectively 16.7% and 22.2.% in the case of the MLP. These two authors are often confused with other
authors. Note that the Quran and Hadith books are attributed without any error (error of 0%).
© Copyright Sayoud.net 2021-2025, All rights reserved. 58
Table 12.9: Identification Error in % using the feature: Character-tetragram, on SAB-1 dataset.
Total
Identifica-
tion error
on the 7
books
The
holy
Quran
book
The
Hadith
book
Aaid’s
book
Abdelkafy’s
book
Alghazali’s
book
Alquaradawi’s
book
Amro-
Khaled’s
book
Date / Century
Ancient
6th
century
Ancient
6th
century
Recent
20th
century
Recent
20th century
Recent
20th century
Recent
20th century
Recent
20th
century
Classifier
Manhatan
centroid
distance
1.05%
0%
0%
0%
0%
0%
11.1%
0%
MLP
classifier*
2.1%
0%
0%
6.25%
16.7%
0%
0%
0%
SMO-SVM
classifier*
3.1%
0%
0%
12.5%
16.7%
0%
0%
0%
Linear
Regression*
2.1%
0%
0%
6.25%
16.7%
0%
0%
0%
*: 500 most frequent features only.
In table 12.9, we can see that the best classifier is Manhattan distance, which gives an error of only
1.05%, the other classifiers present different performances (total identification errors ranging between
2.1% and 3.1). The three authors: Aaid-Alkarni, Abdelkafy and Alquaradawi present some problems of
authorship attribution depending on the choice of the classifier. These two first ones are often confused
with other authors. As previously, we can note that the Quran and Hadith books are attributed without
any error (error of 0%).
Table 12.10: Identification Error in % using the feature: Word, on SAB-1 dataset.
Total
Identifica-
tion error
on the 7
books
The
holy
Quran
book
The
Hadith
book
Aaid’s
book
Abdelkafy’s
book
Alghazali’s
book
Alquaradawi’s
book
Amro-
Khaled’s
book
Date / Century
Ancient
6th
century
Ancient
6th
century
Recent
20th
century
Recent
20th century
Recent
20th century
Recent
20th century
Recent
20th
century
Classifier
Manhatan
centroid
Distance
1.05%
0%
0%
6.25%
0%
0%
0%
0%
MLP
classifier*
1.05%
0%
0%
0%
16.7%
0%
0%
0%
SMO-SVM
classifier*
2.1%
0%
0%
0%
0%
33.3%
0%
0%
Linear
Regression*
2.1%
0%
0%
6.25%
16.7%
0%
0%
0%
*: 500 most frequent features only.
In table 12.10, we can see that the best classifiers are the MLP and Manhattan distance, which give an
error of only 1.05%, the other classifiers present the same performances (total identification errors of
2.1%). The two authors: Aaid-Alkarni and Abdelkafy present some problems of authorship attribution
depending on the choice of the classifier. These two particular authors are often confused with other
authors. Also, as in the previous results, we can note that the Quran and Hadith books are attributed
correctly (error of 0%).
© Copyright Sayoud.net 2021-2025, All rights reserved. 59
Table 12.11: Identification Error in % using the feature: Word Bigram, on SAB-1 dataset.
Total
Identifica-
tion error
on the 7
books
The holy
Quran
book
The
Hadith
book
Aaid’s
book
Abd-
elkafy’s
book
Algha-
zali’s
book
Alquara-
dawi’s
book
Amro-
Khaled’s
book
Date / Century
Ancient
6th
century
Ancient
6th
century
Recent
20th
century
Recent
20th
century
Recent
20th
century
Recent
20th
century
Recent
20th
century
Classifier
Manhatan
centroid
distance
1.05%
0%
0%
0%
0%
3.1%
0%
0%
SMO-
SVM
classifier#
3.1%
0%
0%
12.5%
16.7%
0%
0%
0%
MLP
classifier#
4.2%
0%
0%
12.5%
33.3%
0%
0%
0%
Linear
Regression
#
4.2%
0%
0%
12.5%
16.7%
0%
0%
20%
#: 600 most frequent features only.
In table 12.11, we can see that the best classifier is Manhattan distance, which gives an error of only
1.05%, the other classifiers present different performances (total identification errors ranging between
3.1% and 4.2%). The three authors: Aaid-Alkarni, Abdelkafy and Alghazali present some problems of
authorship attribution depending on the choice of the classifier. Again, these two first ones are often
confused with other authors. Once again, as in the previous results, we can note that the Quran and
Hadith books are attributed without any error (error of 0%).
Note: we notice that Manhatan centroid distance, which is a relatively simple statistical classifier,
outperforms the other machine learning classifiers in many cases. However we do know that these last
ones are usually better than the distance based classifiers especially for the SVM classifier, which is
considered as the state-of-the-art classifier in many research fields. The main possible reason is the low
dimensionality of the training dataset, which usually leads to a weak training process (note that some
books are too small with only 8 or 9 texts per book: this fact makes difficult to get a big training dataset).
12.3.2 Experiments of authorship attribution using fusion techniques
In order to further enhance the authorship attribution performances, two fusion techniques have been
proposed and implemented: the FDF and CDF fusion techniques. We can see in tables 12.12 and 12.13
the corresponding results of those two fusion techniques respectively.
Table 12.12: Error of identification using the feature-based fusion (FDF)
Total Identification
error on the 7 books
The holy
Quran
The
Hadith
book
book
book
book
Amro-
0%
0%
0%
0%
0%
0%
0%
0%
Table 12.13: Error of identification using the classifier-based fusion (CDF)
Total Identification
error on the 7 books
The holy
Quran
The
Hadith
book
book
book
book
Amro-
0%
0%
0%
0%
0%
0%
0%
0%
© Copyright Sayoud.net 2021-2025, All rights reserved. 60
As we can see in tables 12.12 and 12.13, the authorship attribution error is equal to zero for every author.
Te total identification score is 100%, showing the superior performances of the fusion techniques over
the conventional classifiers as expected in theory. This result is very interesting since it shows that a
combination of different features and/or classifiers can lead to high authorship attribution performances.
12.4 Discussion and comments
By observing the different experimental results, we can see that the 7 different books have been
discriminated (let us say) correctly with regards to the writer/author: the corresponding text segments
have been attributed to the correct authors with a small error of identification. Moreover, by using the
fusion approach the attribution error have been reduced to only 0%. This important result shows that the
classical features and classifiers that are usually employed in English and Greek languages got good
results for the Arabic language too and appear to be utilizable for the authorship attribution of texts that
are written in Arabic.
The first conclusion we can state is that the fusion approach is quite interesting in multi-classifier or
multi-feature authorship attribution.
Another important conclusion, one can deduce, is that the two religious books Quran and Hadith appear
to have two different Authors. In fact, all the experiments have shown that the Quran segments and
Hadith segments are identified and classified separately without any attribution error, with or without
the use of fusion.
Now, for a reason of statistics, let us have a look at table 12.1. This table shows the following indications
on the cross-combinations of the authorship identification tests: the total number of possible cross-
combinations tests NtestQH, between Quran segments and Hadith segments, is equal to:
NtestQH = (23 x 4) + (4 x 7) = 92 + 28
=120 different combinations. (12.4)
So, by supposing that at the 121th combination we will get an error of attribution between one Quran
segment and one Hadith segment, then the probability to get an error of attribution would be 1/121=
0.8%.
Since this is only a supposition and not a real attribution error, then we can say that the probability of
authorship confusion between the two religious books is less than 0.8% in the present experimental
conditions.
Furthermore, since there are no cross-errors of attribution between the Quran and Hadith texts (each
other) and more generally, since there was not any cross-error of attribution for the Quran texts or Hadith
texts, with regards to the 6 other investigated books, we can state that these 2 books are completely
different in style each other and also different from all the other investigated books. Consequently, it
appears that the Quran and Hadith should have two different authors or at least two different author
styles.
© Copyright 2021-2025 Sayoud.net. All Rights Reserved. 61
13 Seventh Series of Experiments: Authorship
Discrimination using the Leave-One-Out Validation
In this survey, we employ the Leave-One-Out cross-validation (LOO) in the stylometric analysis of the two
religious books (i.e. Quran and Hadith). In fact, although conventional classification approaches were
widely employed in the literature, they still present a lack of statistical consistency. The LOO and LTO
cross-validation techniques consist in different experiments of authorship attribution that are carried out in
a rotating manner, excluding every time one or two new samples.
Hence, two experiments of authorship classification are made on the two religious books, which are
expected to be fair and significant. The books are segmented into distinct text segments of 2900 tokens
each, and the used features are composed of character-tetragrams, which are known to be quite efficient in
stylometry.
The proposed discrimination approach is based on the Leave-One-Out (LOO) and Leave-Two-Out (LTO)
cross-validation techniques using a Support Vector Machines (SVM) based classifier.
13.1 The Leave-One-Out Method
The Leave-One-Out Method is a jackknife method for evaluating the classification accuracy (Vehtari 2016).
It was proposed by Lachenbruch in 1967 (Lachenbruch, 1967). His approach was based on discriminant
analysis; it has been named the leave-one-out (LOO) method (Huberty, 1994). This technique has two steps:
-First, the template is built in the samples with one observation removed,
-Then the resulting estimate parameters (of the training) are used to classify the single removed
observation.
The main process is repeated M times so that each observation was removed and classified once (see Figure
13.1), where M represents the number of samples (Kroopnick et al., 2010).
Eventually, the proposed measure of good classification is given by the number of times that the removed
observation was correctly classified (Huberty, 1994) (Kroopnick et al., 2010).
To evaluate the LOO method, Lachenbruch conducted a small Monte Carlo simulation with 300 replications
(Kroopnick et al., 2010).
© Copyright 2021-2025 Sayoud.net. All Rights Reserved. 62
Figure 13.1.a.
Figure 13.1.b. The Leave-One-Out algorithm applied to Sample 1 (start of the algorithm).
Sample
1
Sample
2
Sample
3
Sample
4
Sample
N
Sample
1
Sample
2
Sample
3
Sample
4
Sample
N
Training Model
© Copyright 2021-2025 Sayoud.net. All Rights Reserved. 63
Figure 13.1.c. The Leave-One-Out algorithm applied to Sample 2 and moving to the next sample.
Figure 13.1.d. The Leave-One-Out algorithm applied to Sample 3 and moving to the next sample.
Sample
1
Sample
2
Sample
3
Sample
4
Sample
N
Sample
1
Sample
2
Sample
3
Sample
4
Sample
N
Training Model
Training Model
© Copyright 2021-2025 Sayoud.net. All Rights Reserved. 64
Figure 13.1.e. The Leave-One-Out algorithm applied to Sample 4 and moving to the next sample.
Figure 13.1.f. The Leave-One-Out algorithm applied to Sample N (end of the algorithm).
Similarly, the Leave-Two-Out is an extension of the Leave-One-Out algorithm, where instead of
excluding only one sample, we should exclude two samples simultaneously. It is referred to by the
abbreviation LTO.
Sample
1
Sample
2
Sample
3
Sample
4
Sample
N
Sample
1
Sample
2
Sample
3
Sample
4
Sample
N
Training Model
Training Model
© Copyright 2021-2025 Sayoud.net. All Rights Reserved. 65
13.2 On the Choice of the Classifier and Feature
Concerning the task of authorship attribution (Ouamour et al., 2013), we previously investigated the
authorship of several short historical texts that were written by ten ancient Arabic travelers: called AAAT
dataset. Several experiments of authorship attribution were conducted on those Arabic texts, by using
seven different classifiers, namely: Manhattan distance, Cosine distance, Stamatatos distance, Camberra
distance, Multi-Layer Perceptron (MLP), Sequential Minimal Optimization based Support Vector
Machine (SMO-SVM) and Linear Regression. The results showed that the best performances of
authorship attribution were given by the SVM, which outperformed the other investigated classifiers.
For this reason, and knowing the good performances of the SVM in discrimination, we have decided to
use this classifier for the task of authorship discrimination.
Concerning the features, in the present investigation, we have used the character-tetragrams.
13.3 Text Segmentation
A text segmentation is applied in order to construct individual documents with the same size. In fact,
when comparing two books with different sizes, it is difficult to know if a specific part of the book is
similar to another one or different. That is why a smart segmentation has been proposed and applied to the
different books.
The sizes of the segments are more or less in the same range: we obtained 29 different text segments for
the Quran and 8 different text segments for the Hadith, with approximately the same size. So, we got 37
different text segments of about 2900 words each in the whole dataset. Our chosen configuration seems to
be correct and suitable to the different AA experiments.
The segmented dataset is decomposed into 2 rotating Leave-K-Out parts: the training part containing all
the text samples except K (i.e. one or two), and the testing part consisting in the removed ones.
13.4 Experiments of AA using the LOO and LTO techniques
We recall that we used the character-tetragram feature by keeping only the 500 most frequent features, and
employed the SMO-based SVM classifier.
Since there are 37 samples, we will also have 37 experiments of rotating classification, where in every
experiment one sample (or two) is removed and put in testing set, in order to be identified through the
remaining samples that represent the training model.
In the following table, we represent the scores of good classification obtained by the first series of
experiments of LOO, which corresponds to the 37 LOO cross validation experiments.
© Copyright 2021-2025 Sayoud.net. All Rights Reserved. 66
Table 16.1: Results of AA using the LOO technique
Experiment Number
Tested document
Accuracy
1.
Q1
100%
2.
Q2
100%
3.
Q3
100%
4.
Q4
100%
5.
Q5
100%
6.
Q6
100%
7.
Q7
100%
8.
Q8
100%
9.
Q9
100%
10.
Q10
100%
11.
Q11
100%
12.
Q12
100%
13.
Q13
100%
14.
Q14
100%
15.
Q15
100%
16.
Q16
100%
17.
Q17
100%
18.
Q18
100%
19.
Q19
100%
20.
Q20
100%
21.
Q21
100%
22.
Q22
100%
23.
Q23
100%
24.
Q24
100%
25.
Q25
100%
26.
Q26
100%
27.
Q27
100%
28.
Q28
100%
29.
Q29
100%
30.
H1
100%
31.
H2
100%
32.
H3
100%
33.
H4
100%
34.
H5
100%
35.
H6
100%
36.
H7
100%
37.
H8
100%
In the following table, we represent the scores of good classification obtained by the second series of
experiments of LTO, which corresponds to the 19 LTO cross validation experiments.
© Copyright 2021-2025 Sayoud.net. All Rights Reserved. 67
Table 16.2: Results of AA using the LTO technique
Experiment Number
Tested documents
Accuracy
1.
Q1,2
100%
2.
Q3,4
100%
3.
Q5,6
100%
4.
Q7,8
100%
5.
Q9,10
100%
6.
Q11,12
100%
7.
Q13,14
100%
8.
Q15,16
100%
9.
Q17,18
100%
10.
Q19,20
100%
11.
Q21,22
100%
12.
Q23,24
100%
13.
Q25,26
100%
14.
Q27,28
100%
15.
Q29,30
100%
16.
H1,2
100%
17.
H3,4
100%
18.
H5,6
100%
19.
H7,8
100%
It will be now interesting to compute the average accuracy, corresponding to the overall performances of
classification. This entity can be evaluated by using equation 16.1.
Average Accuracy =
(16.1)
where N represents the number of cross-validation experiments (denoted by CrossVal).
According to table 16.1, the average accuracy of all LOO experiments is equal to 100%.
And, according to table 16.2, the average accuracy of all LTO experiments is equal to 100%.
13.5 Discussion
The two ancient religious books (i.e. Quran and Hadith) have been analysed by a discriminative authorship
analysis using the Leave-One-Out and Leave-Two-Out validation techniques. The features consist in
character-tetragrams, while the used classification is based on an SMO-SVM classifier. The dataset is
composed of 37 text documents, where the size of a single segment is about 2900 tokens.
As we could see in the results section, the accuracy of every cross-validation step, for all LOO and LTO
experiments, was 100%, leading to an average cross-validation score of 100% too.
From these results, one can deduce the following important conclusions:
- Firstly, the two books Quran and Hadith appear to have two different author styles;
- The segments of every book are quite similar in terms of style within a single book;
- The LOO and LTO cross validation techniques have shown that this result (discrimination score of
100%) is quite significant, since the same score has been obtained 37+19 times, namely 56 times
during the tests of cross-validation and with different configurations.
Consequently and according to this investigation, it is clear that the two ancient books: Quran and Hadith
possess two different styles and should probably come from two different Authors. Finally, and once again,
one can deduce that the Quran could not be written by the Prophet.
© Copyright 2021-2025 Sayoud.net. All Rights Reserved. 68
14 Eighth Series of Experiments: Authorship
Discrimination based on Gaussianity and
Interpolability
It is well known that every physical phenomenon in the universe respects some specific rules, as we can
scientifically observe and measure. For instance, the Gaussianity rule characterizes every physical
On the other hand, the Interpolability can be
noticed for almost every discrete curve representing a natural physical phenomenon. Those two rules are
respected by a wide variety of data present in the Universe or let us say simply in our daily life.
In the case of textual data, the two rules should theoretically be respected, and indeed, they have been
verified during this investigation.
However, in the case of the holy Quran, this study shows that the Gaussianity and interpolability of the data
curve do not appear to be respected. Furthermore, we notice an inexplicable and strange statistical structure
in the holy book, without any (prior) scientific interpretation.
14.1 Introduction
Most of existing natural data obey a certain set of physical rules that seem to be quite simple, mathematically
speaking. For instance, the gravity equation is basically simple (W=m.g); the energy equation is also simple
(E=mc²) and the electric voltage equation is further simpler (U=RI). This mathematical simplicity in most
of the natural or physical data is very often verified.
mean of a sufficiently large number of iterates of independent random variables, each with a well-defined
expected value and well-defined variance, will be approximately normally distributed (i.e. Gaussian
distribution), regardless of the underlying distribution [Siegrist, 2016] [Rice, 1995]. That is, suppose that a
sample is obtained containing a large number of observations, each observation being randomly generated
in a way that does not depend on the values of the other observations, and that the arithmetic average of the
observed values is computed. If this procedure is performed many times, the central limit theorem says that
the computed values of the average will be distributed according to the normal distribution (i.e. Gaussian
distribution) (Contributors, 2015).
The central limit theorem has an interesting history. The first version of this theorem was postulated by the
French mathematician De Moivre who, in a remarkable article published in 1733, used the normal
distribution to approximate the distribution of the number of heads resulting from many tosses of a fair coin.
This finding was far ahead of its time, and was nearly forgotten until the French mathematician Laplace
as published in
1812. Laplace expanded De Moivre's finding by approximating the binomial distribution with the normal
distribution. But as with De Moivre, Laplace's finding received little attention in his own time. It was until
the end of nineteenth century that the importance of the central limit theorem was discerned, when, in 1901,
the Russian mathematician Aleksandr Lyapunov defined it in general terms and proved precisely how it
worked mathematically. Nowadays, the central limit theorem is considered to be the sovereign of
probability theory (Contributors, 2015).
Sir Francis Galton described the Central Limit Theorem as (Galton, 1889)
in complete self-effacement, amidst the wildest confusion. The larger the mob, and the greater the apparent
anarchy, the more perfect is its sway. It is the supreme law of Unreason. Whenever a large sample of chaotic
elements are taken in hand and marshaled in the order of their magnitude, an unsuspected and most beautiful
form of regularity proves to have been latent all along>> (Contributors, 2015).
© Copyright 2021-2025 Sayoud.net. All Rights Reserved. 69
Another important aspect is the natural continuity of every physical or natural phenomenon. In other words,
the graphical representation of a measured physical data should present a certain continuity and regularity
(i.e. the curve shape respects some well-known graphical models). Again, taking discrete samples from that
measured physical data, will lead to a discrete curve from which it should be possible to interpolate and fit
with usual interpolation or fitting functions.
Hence, it could be seen why Michael Whiteman stated (Whiteman, 1967)
continuity and gradient, which are applicable to conceptually defined curves, and thence are applicable
approximately to physical. curves. Thus, in any particular case a physical curve may be tested for continuity,
and an approximate measure of its gradient may be found. Likewise a physical trajectory does not consist
of isolated events. Nevertheless, by selecting points, the exact concept of velocity may be applied so as to
(Whiteman, 1967).
Now, by considering some natural/physical curves present in the universe, we observe continuity in the
graphical representation of the measured dimension (continuity of the dimension). Moreover, we should
even find some continuity in its first derivative (continuity of the variation). For instance, let us observe the
temperature curve of the weather. Not only, the temperature should vary continuously and smoothly, but
also its derivative does so, and this with an extreme respect of the physical and natural well-known laws.
Again, by observing the natural curves present in real life, one remark that the form of the curves is quite
identifiable by a simple visual observation (as experts do), and the curves are easily fitted by usual
mathematical functions (e.g. Gaussian, Linear, Polynomial, Sinusoidal, Exponential, etc.).
For concreteness, if we take the text data for instance, we may remark that it is composed of several
i.e. vocabulary), several numbers (e.g. 1, 2,
3 etc.) and so on.
That is, if a large amount of textual data is analyzed by computing the frequency of some features, we
should usually retrieve a Gaussian distribution of the data, when the features are represented from the most
frequent to the least frequent.
As it can be seen in the next section, this particularity has been checked with 7 different books by taking
Contrariwise, for the holy Quran those mathematical rules do not seem to be respected, without any possible
interpretation. Moreover, by analyzing another feature (i.e. Number citation frequency), we strangely
noticed that this last feature does not respect the previous laws either, while for the case of the Hadith book,
the mathematical laws are well respected for all those features.
14.2 Fitting and interpolation definitions
Given a set of data that results from an experiment or from a physical scenario, we show (in Mathematics)
that there is some function that passes through the data points and optimally represents the area of interest
at all present and absent points.
With interpolation (Milne, 2012), we look for a function that allows us to approximate the values between
the original data points [William Edmund Milne]. The interpolating function should pass exactly through
the original data points. In our experiments, we chose two different techniques: the Piecewise cubic Hermite
interpolation (PCHIP), which preserves the monotonicity and shape of the data and the Bezier interpolation
technique, which preserves the curve shape in a graphical way.
On the other hand, with curve fitting (Milne, 2012), we simply want a function that represents a good fit to
the original data points with a minimum estimation error. With curve fitting the approximating function
does not have to pass through the original data points. It should respect the overall data with the best possible
fitting, and respect the chosen mathematical expression type as well.
© Copyright 2021-2025 Sayoud.net. All Rights Reserved. 70
For instance, in table 14.1, we have represented four columns : the first one represents the sample order, the
second one represents the original data corresponding the real samples values, the third column represents
the interpolation based values and the fourth one represents the fitting based values.
Table 14.1. Example of Interpolation and Fitting.
Sample
Original Samples
Interpolation
Fitting
1
1.00
1.00
1.07
2
absent
2.01
2.04
3
3.00
3.00
2.99
4
absent
3.97
3.98
5
5.00
5.00
5.03
14.3 Investigation on the Word-Length frequency
14.3.1. Definition of the Word Length Frequency (WLF)
Since the first part of our experiment concerns the Word Length Frequency (WLF) (H. Sayoud, 2012), we
think that it could be useful to define some technical terms employed in this work:
-The "Word Length" represents the number of letters composing the word.
-
composed of n letters each, present in the text.
14.3.2. Graphical representation of the Word Length Frequency
In this section we will graphically represent the WLF of the two books : holy Quran and Hadith.
Furthermore, we will represent the WLF of 6 other books written by 6 different authors to make a general
comparison between their features.
© Copyright 2021-2025 Sayoud.net. All Rights Reserved. 71
Figure 14.1: Word length frequency of the Quran (in red) and Hadith (in blue), obtained with Bezier
interpolation. We can notice that only the Hadith curve presents a Gaussian (or log-normal) shape.
As we can see from figure 14.1, in a visual point of view, the Hadith curve (in blue) respect the Guassianity,
whereas the Quran curve (in red) does not seem to respect any Guassianity or at least any semi-Gaussianity
(i.e. Gaussianity in one side).
By the term Gaussianity, we mean a smooth bell curve, which more or less resemble to a mono-Gaussian form
in one or both sides.
We also visually notice that the ideal curve that could, maybe, ensure a certain Guassianity for the Quran is
disconnected from the real Quran curve at abscises 3 and 7.
Due to that strange observation, we have decided to test if the Guassianity is also respected with other long
Arabic texts (eg. testing other books/authors) or not. So, we have basically drawn the WLFs of 6 other
books written by 6 different authors to see if there is any possible Gussianity or at least a log-normal shape.
Hence, several experiments of Word length frequency have been conducted on the holy Quran and the books
of 7 other religious Authors, namely: the Prophet (PBUH), Dr Abd AlKafy, Dr Amro Khaled, Dr Al
Ghazali, Dr Al Arifi, Dr Al Qarqdqwi and Dr Hassan. The 6 last ones represent contemporary authors from
the 20th and 21th century, for which the main topic is also religion. Results are represented in figure 2.
0 1 2 3 4 5 6 7 8 9 10
0
5
10
15
20
25
0
5
10
15
20
25
0 1 2 3 4 5 6 7 8 9 10
Frequency
Quran
Hadith
© Copyright 2021-2025 Sayoud.net. All Rights Reserved. 72
Figure 14.2: The word length frequency of 8 different books: the Quran (in red), Hadith (in blue) and 6
other books written by 6 different contemporary authors, obtained with Bezier smoothing. We notice
that all the curves, except the Quran, present a Gaussian or log-normal shape, at least for one of the two
sides.
1
23
4
5
6
78910
1
2
34
5
6
7
8910
0
10
20
30
40
50
60
70
80
90
0246810 12
Quran
Hadith
AmroKhaled
Alquaradawi
Alghazali
Abdelkafy
Alarifi
Hassan
© Copyright 2021-2025 Sayoud.net. All Rights Reserved. 73
By observing figure 14.2, we notice that all the seven WLF curves present a certain gaussianity (for at least one
side), except the Quran one, which has a strange graphical shape that does not seem to respect any form of
gaussianity.
In a separate representation, we can see the Quran WLF with more details (see figure 14.3), where it clearly
appears that it does not respect any Gaussianity or Interpolability. Moreover, we notice that the general form of
the curve is quite strange (not familiar).
Figure 14.3: The word length frequency of the holy Quran, obtained with PCHIP interpolation: we notice that
the curve shape is not Gaussian and not log-normal either, in neither the left nor the right side. We also
remark that it is not interpolable with conventional interpolation functions. In fact, two exceptions, precisely
in 3 and 7, make the Gaussianity and Interpolability not respected.
In the previous figures (14.1 and 14.3), a strange form is noticed for the Quran WLF and a question could be
asked then: is it the result of a mixture of two (or more) styles? In other terms, does the holy Quran result from
a mixture of several authors? That question could be statistically stated as: is the Quran representation multi-
gaussian?
To answer that question, we simulated several text mixtures and computed the corresponding approximated
WLFs.
In the first experiment, we simulated the text mixture of the two authors: Dr Hassan and Dr Amro-Khaled. See
figure 14.4.a.
1 2 3 4 5 6 7 8 9 10
0
2
4
6
8
10
12
14
16
18
20
Word length
Frequency
Quran
© Copyright 2021-2025 Sayoud.net. All Rights Reserved. 74
Figure 14.4.a: WLF of a simulated mixture between the texts of two different authors: Dr Hassan and Dr
Amro-Khaled.
In the second experiment, we simulated the text mixture of the two authors: Dr Abd-AlKafy and Dr Amro-
Khaled. See figure 14.4.b.
Figure 14.4.b: WLF of a simulated mixture between the texts of two different authors: Dr Abd-AlKafy and Dr
Amro-Khaled.
In the third experiment, we simulated the text mixture of the two authors: Dr Al-Ghazali and Dr Al-Qaradawi.
See figure 14.4.c.
000
005
010
015
020
025
0246810 12
Mixture of Hassan+Amro-Khaled
Series1
000
005
010
015
020
025
0246810 12
Mixture of Abd AlKafy+Amro-Khaled
Series1
© Copyright 2021-2025 Sayoud.net. All Rights Reserved. 75
Figure 14.4.c: WLF of a simulated mixture between the texts of two different authors: Dr Al-Ghazali and Dr
Al-Qaradawi
In the fourth experiment, we simulated the text mixture of the two authors: Dr Al-Arifi and the Prophet (Pbuh).
See figure 14.4.d.
Figure 14.4.d: WLF of a simulated mixture between the texts of two different authors: Dr Al-Arifi and the
Prophet (Pbuh).
In the fifth experiment, we simulated the text mixture of 7 different authors: Dr Hassan, Dr Alarifi, Dr Alkarny,
Dr Abdelkafy, Dr Alghazali, Dr Alquaradawi and Dr AmroKhaled. See figure 14.4.e.
000
005
010
015
020
025
0246810 12
Mixture of AlGhazali+AlQaradawi
Series1
000
005
010
015
020
025
0246810 12
Mixture of Alarifi + Hadith
Series1
© Copyright 2021-2025 Sayoud.net. All Rights Reserved. 76
Figure 14.4.e: WLF of a simulated mixture between the texts of 7 different authors: Dr Hassan, Dr Alarifi, Dr
Alkarny, Dr Abdelkafy, Dr Alghazali, Dr Alquaradawi and Dr Amro-Khaled.
As we can see in all the simulated texts/authors mixture, we got always a Gaussian (on at least one side), and
we do not observe any division of the curve into two or multiple Gaussian as one could expect.
This fact shows that the hypothesis of multiple styles in the holy Quran is excluded, since that phenomenon has
not been noticed in the previous simulations.
So, what could be then the reason of that strange unexplained form?
In the opinion of the author, there is no classic explanation for that fact, except that the Quran should be the
work of the Creator who made his holy scripture above the classic rules of statistics and mathematics.
14.3.3. Hadith model interpolated with Gaussian fitting
From the previous results, showing that the Hadith should respect a certain Gaussianity and Interpolability, we
performed a computation of a Gaussian curve in a form given by equation 14.1, and optimized it to get the lowest
error possible (i.e. optimal coefficients for the best fitting).
f(x) = a1*exp(-((x-b1)/c1)^2) (14.1)
The obtained results are given below:
Parameters
a1 = 23.61
b1 = 3.512
c1 = 2.497
Goodness of fit:
SSE: 2.263
R-square: 0.9969
RMSE: 0.5686
The resulting fitted curve is represented in figure 14.5.
0.00
5.00
10.00
15.00
20.00
25.00
0246810 12
Mixture of 7 authors
Series1
© Copyright 2021-2025 Sayoud.net. All Rights Reserved. 77
Figure 14.5: The word length frequency of the Hadith , obtained with a Gaussian fitting: we notice that the
Gaussian curve fits the data with a quite good precision.
Observation
From the previous results, we notice that the Hadith and all the cited books obey to the law of Gaussianity and
Interpolability, except the holy Quran, which does not respect any of these classical laws. That is, we do not
understand why this particular exception is noticed for the holy scripture. Again, the only interpretation, one
can give, is that the concerned book should have a mysterious origin.
14.4 Investigation on Numbers citation frequency
In this second investigation, we consider the citation of numbers in the text, such as one, two, three, etc.
However, those numbers are sorted from the most frequent to the least one. For concreteness, if the numbers 1,
2 and 3 have the following frequencies 10%, 15% and 12% respectively, then they will be sorted into the
following sequence: 2 (1st number), 3 (second number) and 1 (3rd number). That scheme makes the
representation curve monotone (decreasing) and easier to interpolate.
On the other hand, only numbers that are cited at least more than 5 times are considered, for a purpose of
consistency. Consequently and in practice, only the 6 or 7 most frequent numbers are kept in the graphical
representation.
1 2 3 4 5 6 7 8 9 10
0
5
10
15
20
Word length
Frequency
Hadith
© Copyright 2021-2025 Sayoud.net. All Rights Reserved. 78
Figure 14.6: Number citation in the holy Quran (sorted from the most frequent to the least frequent). The
curve is obtained by Bezier interpolation.
Figure 14.7: Number citation in the Hadith (sorted from the most frequent to the least frequent). The curve is
obtained by Bezier interpolation.
0.00
0.05
0.10
0.15
0.20
0.25
0.30
0.35
0.40
0123456
Frequency
Rank of the most cited numbers
Quran
0.00
0.05
0.10
0.15
0.20
0.25
0.30
0.35
0 1 2 3 4 5 6
Frequency
Rank of the most cited numbers
Hadith
© Copyright 2021-2025 Sayoud.net. All Rights Reserved. 79
Figure 14.8: Quran number frequency: there is no Gaussianity and no Interpolability either. Once again, the
mathematically speaking. The curve is obtained by PCHIP interpolation.
By observing figure 14.8, we remark that the curve presents different slopes at each segment located between
two successive points. The general curve is quite strange and unfamiliar, mathematically speaking. Once again,
there is no Gaussianity and no Interpolability either. Furthermore, the curve does not resemble to any known
mathematical or physical shape. Also, we notice that there exists a pseudo-horizontal segment between the 2nd
and 3rd numbers and between the 6th and 7th ones.
Hadith model interpolated with Exponential fitting f(x) for the Number frequency
As in the previous investigation, and due to the fact that the Hadith curve appears to respect a certain Gaussianity
and Interpolability (i.e. visually), we performed a computation of an exponential curve in a form given by equation
14.2, and optimized it to get the lowest error possible.
f(x) = a*exp(b*x) + c*exp(d*x) (14.2)
The obtained results are given below:
Parameters
a = 61.85
1 2 3 4 5 6 7
0
10
20
30
40
50
Number
Frequency
Quran
© Copyright 2021-2025 Sayoud.net. All Rights Reserved. 80
b = -0.9579
c = 8.81
d = -0.007993
Goodness of fit:
SSE: 0.472
R-square: 0.999
RMSE: 0.3967
The resulting fitted curve is represented in figure 14.9.
Figure 14.9: Hadith number frequency: Once again, the overall curve seems to follow a partial Gaussianity
shape and the Interpolability is possible for every point, by an exponential polynomial, as we can see in the
corresponding fitting equation. The curve is obtained with exponential fitting.
Observation
As in the previous investigation, we notice again that the Hadith obeys to the law of Guassianity and
number citation presents a complex curve with no Guassianity or Interpolability either. This fact, once again,
proposes that the holy scripture should have a mysterious origin.
1 2 3 4 5 6 7
10
15
20
25
30
Number
Frequency
Hadith
© Copyright 2021-2025 Sayoud.net. All Rights Reserved. 81
14.5 Conclusion and Discussion
An investigation of Gaussianity and Interpolability has been conducted on the Holy Quran, in order to see if it
respects, as the other human books, the physical properties of Gaussianity and Interpolability.
A summary of the different curves is given in the following table (table 14.2).
Table 14.2. Comparison between the Quran and Hadith curves (Bezier interpolation is used in both books).
Holy Quran
Hadith
Sorted Number citation frequency
Word length frequency
As a comparison with other books, we positively verified that the Hadith book does obey to the Gaussianity
and Interpolability rules for both Word length frequency and Number citation. Similarly, 6 other books written
by different human authors have been analyzed and experimented in the same manner. Once again, those 6
different books appear to obey perfectly to the Gaussianity and Interpolability rules for the word length
frequency.
On the other hand, and contrariwise, we strangely noticed that the holy Quran does not respect those rules for
the word length frequency and for the number citation frequency either.
condition (at least to the knowledge of the author). However in the case of the holy Quran, neither the
Gaussianity nor the continuity of the curve evolution (second derivative) is respected. This fact proposes that
the holy Quran could not be a human invention but probably the work of a Superior Non-Human Intelligence
who is beyond the prescribed rules and who does not respect any of the well known physical properties.
Consequently, we may deduce two important facts:
0
5
10
15
20
25
30
35
40
0 2 4 6
0
5
10
15
20
25
30
35
40
0 2 4 6
0
5
10
15
20
25
0 1 2 3 4 5 6 7 8 9 10
0
5
10
15
20
25
012345678910
© Copyright 2021-2025 Sayoud.net. All Rights Reserved. 82
- Firstly, the two investigated books: Quran and Hadith are quite different in terms of textual structure
statistics, which leads to the conclusion that the two corresponding Authors should be different;
- Secondly and more strangely, we do not see any possible human origin for the holy Scripture. Hence,
the hypothesis of a Divine origin, for the holy Quran, is widely supported by the result of this
investigation.
Finally, the present paper is a pure Statistical/ Computational-Linguistic investigation regardless of the religious
aspect of the studied books. It tries only to bring a new scientific discovery, which was hidden and unknown
before, to the scientific community.
© Copyright 2021-2025 Sayoud.net. All Rights Reserved. 83
15 Ninth Series of Experiments: A Mysterious Numerical
Structure in the Quran making it different from other
Human Books
In this investigation, a mixed linguistic-statistical-numerical analysis is performed on the text of the holy Quran
in order to look for any possible presence of hidden numerical structure. In our case, we focused our interest on
number seven, which seems to have a long historical and religious presence in the holy book. This task could
be seen as a pure statistical analysis using some simplified linguistic rules and numerical properties to check
the possibility of a Divine origin (i.e. beyond the human capacities) of the book.
The dataset consists of the 2 forms of the Quran: once with diacritics and another time without diacritics. The
first text is considered as our reference book since it consists in the authentic Quran that is preserved by the
Muslim religious community and respecting the Saudi Quran edition (narrated by Hafs). The second one is
similar to the first book, except that we have removed the diacritics to speed up the word search process.
Several simplified experiments of computational linguistics, which are based on the multiplicity by seven, are
performed and commented on the Holy Scripture. Results are outstanding and surprising: in fact, the seven-
based structure has been discovered in many parts of the holy Quran. This result shows that the possibility of a
human origin for such a fascinating structure is quite impossible. In fact, the advanced scientific structure of
the Quran antedates the period of the Prophet by many centuries.
15.1 Motivation based on the citation of number 7 in the holy Quran
The history of numerical structure began in the 70s with some discoveries on number 19. Later on, very few
(Yakub, 2008) and recently (2002-2008) several
numerical evidences based on number 7 were reported by Kaheel (Kaheel, 2015). These last results concerned
the manipulation and concatenation of some numbers related to the holy Quran, which in many cases, produces
a multiple of 7.
By reading the available documentation on the matter we notice that there exist a sort of hidden numerical
organisation inside the Holy Scripture, which could be seen as a watermarking (Cox, 2007). For instance, let
us look at the following watermarked English text: « So I do let him accept a great space!! Yes its ».
In this example, we do not care about the semantic meaning of the text, it is not our objective, but if we
decompose this text into segments of 7 characters each, we will get the following table (table 15.1). Now, again
if we keep only the letters located at the 1st diagonal of the table (in red), we will get the following word:
« Secret », which was really embedded in the precedent text but in a hidden form.
Table 15.1: Example of text watermarking.
S
o
-
I
-
d
o
l
e
t
-
h
i
m
a
c
c
e
p
t
-
a
-
g
r
e
a
t
s
p
a
c
e
!
!
y
e
s
-
i
t
s
Watermark
S
e
c
r
e
t
© Copyright 2021-2025 Sayoud.net. All Rights Reserved. 84
Such techniques of watermarking (Cox, 2007) are relatively recent and have been proposed during the last
century only.
Now, if such a watermarking does exist in the holy scripture, it is more likely that this signature should be more
sophisticated and much more powerful, since it is supposed to belong to a super-intelligent Author (Allah).
This reason was so motivating that it prompted us to conduct a thorough mixed linguistic-statistical-numerical
analysis on the holy Quran and especially on number 7.
15.2 Citation of number 7 in the holy Quran
The reader of the holy book can easily notice that the number seven is very often used and in many
circumstances too, as we can see in the following cases:
1. The number of heavens (as described in the Quran 65:12) is 7 [Holy Quran]
2. Number 7 is the first number stated in the Quran (Kaheel, 2015)
3. Number 7 is the most repeated number after number 1 (Kaheel, 2015) and (H. Sayoud, 2012). See
figure 15.1.
Figure 15.1: Number citation frequency (Statistics made by H. Sayoud in 2012) (H. Sayoud, 2012)
4. Al-Fatihah (the opening of the Quran) is composed of 21 alphabets (=3x7): a multiple of 7 (Kaheel,
2015)
5. Al-Fatihah (the opening of the Quran) is composed of 7 verses (Kaheel, 2015)
6. As described in the Quran, the Hell possess 7 doors (ref. 15:44) [Holy Quran]
7. The number of alphabets used as Initials (disconnected letters) is 14 (=2x7): a multiple of 7 (Kaheel,
2015)
8. 7×4): a multiple of 7.
9. The Prophet (PBUH) lived for 63 years (=7x9): a multiple of 7 (Kaheel, 2015)
10. During the pilgrimage, Muslims turn around the Kaaba 7 times [Al-Bukhari Hadith 854]
11. Also, during the pilgrimage, Muslims walk between the two hills of Safa and Marwa 7 times [Al-
Bukhari Hadith 854]
© Copyright 2021-2025 Sayoud.net. All Rights Reserved. 85
12. Also, during the pilgrimage, as stated in the Quran, for special circumstances in the pilgrimage, it
is required to fast 7 days when returning back [Holy Quran]
13. Once again, during the pilgrimage, the pilgrims throw 7 pebbles at the 'Jamrat Al-Aqabah' [Al-
Bukhari Hadith 854]
14. In the verse 12:43, the Quran describes the history of Youssof (Joseph) and the famous dream about
the 7 cows [Holy Quran]
15. In the Islamic religion, children are required to begin their prayer from the age of 7 [Al-Bukhari
Hadith 854]
16. In the verse 69:7, the Quran describes the history of Punishment of Aad during 7 nights [Holy
Quran]
17. The example of those
who spend their monies in the cause of Allah is that of a grain that produces seven (7) spikes...
[Holy Quran]
18. In the verse 31:27 , the Quran speaks about the unlimited/numerous words of Allah and cite as
example the « 7 seas [Holy Quran]
19. The verse 15 :87 speaks about the 7 repeated verses : « And We have certainly given you, (O
Muhammad), seven of the often repeated (verses) and the great Qur'an » [Holy Quran]
All these examples show that it should exist an enigma about number seven, but what could be the secret? This
question prompted us to analyse the holy book and try to see if we could retrieve other statistical facts related
to that number.
15.3 New Statistical Evidences based on Number Seven
Trying to respond to the previous question, we made several experiments of statistical word analysis on the
Holy Scripture (Certified Saudi Version of the holy Quran, narrated by Hafs). So, basically, we first selected
the most important words in term of occurrence frequency (most employed words), and then completed that
list by some potential keywords that could have a relationship with any possible numerical evidence, such as
Results were interesting and sometimes surprising, as we can see in the following discovered cases:
1. One of the most frequent words in the holy Quran is the name of God «
». This Divine name has
3 morphological forms in the holy Quran: a) «
» spelled « Allahee », b) «
» spelled «Allahu»
and c) «
» spelled « Allaha ». We present the occurrences of those 3 words as follows:
1.a The name of God «
» spelled « Allahee » has a total occurrence number of 1239=7x3x59
(multiple of 7) if we include all the 112 bismala(s) introducing the different chapters. Again, if we
do not count those 112 bismala(s), the new count becomes 1127=7x7x23, which is still multiple
of 7 (double multiplicity by 7: 7x7). -S- (Ref. Statistics made in April 2015).
1.b The name of God «
» spelled «Allahu» has a total occurrence number of 980=7x7x20, which is a
multiple of 7 (double multiplicity by 7: 7x7), by either keeping the 112 bismallah or not. -S- (Ref.
Statistics made in April 2015).
© Copyright 2021-2025 Sayoud.net. All Rights Reserved. 86
1.c The third name of God «
» spelled « Allaha » has a total occurrence number of 592=37x16
(multiple of 37). In this case number 37 seems to be particularly interesting since the expression
There is no God but Allah
of the Islam, has also an occurrence number of exactly 37 in the holy Quran. -S- (Ref. Statistics
made in April 2015).
2. Again, concerning the words that are very frequently used, we strangely noticed that the total
occurrence number of the coordination conjunction « », meaning « OR », in the holy Quran, is
280=7x40 (Ref. Statistics made in April 2016).
3. In the holy Quran, it is clearly stated that there are 7 heavens. Amazingly, the term « Heavens »
«
» is cited 189 times =7*27=3*3*3*7 (Ref. Statistics made in April 2016).
4. We strangely noticed that the total occurrence number of the word/verb « », meaning
« Number », in the holy Quran, is exactly 7. This word is extremely interesting since it is the most
important keyword in such numerical analysis. In fact, this was one of my first discoveries making
me so surprised that I pushed farther my analysis to other potential keywords in the holy book. So,
has been repeated exactly 7 times in the Quran? -S- (Ref. Statistics made in April 2015).
5. We strangely noticed that the total occurrence number of the word « », meaning « Words», in
the holy Quran, is exactly 14 =2x7 (multiple of 7). This word is also particularly interesting since
it is one of the most important keywords in our investigation. In fact, it seems like an enigma where
the central point is focused on the repetition of words, and we strangely notice that the token
-S- (Ref. Statistics made in April
2015).
6. We strangely noticed that the total occurrence number of the verb « », meaning « To Count/
Make Statistics », in the holy Quran, is exactly 7. As previously, this word is semantically pertinent
-S- (Ref. Statistics made in April 2015).
7. We strangely noticed that the total occurrence number of the word «
», meaning « The Quran.
», in the holy Quran, is exactly = x7 (double multiplicity by 7). This word represents the central
keyword of our investigation, since it represents the dataset (the book) where all the numerical
but even more: it is exactly equal to 7x7, namely a strong relationship with number 7. Could this
be really a coinci
=3x23. Furthermore, note that the Quran had been sent down for a
period of 23 years: a strange coincidence again. -S- (Ref. Statistics made in April 2015).
8. We strangely noticed that the total occurrence number of the expression « », meaning
« Those are the signs/evidences of the book », which make reference to the mysterious
disconnected letters (sort of acronyms) [Sayoud 2013] in the holy Quran, is exactly7. As stated in
its translation (those are the signs/evidences of the book), it is likely that this expression involves
a hidden secret or proof in the holy Quran with regards to the disconnected letters. In fact, why is
this expression put just after the mysterious disconnected letters? And what sort of sign or evidence
should it provide? And finally, why is it repeated seven times? Sincerely, one cannot say that this
is a pure hazard. -S- (Ref. Statistics made in April 2015).
© Copyright 2021-2025 Sayoud.net. All Rights Reserved. 87
9. We strangely noticed that the total occurrence number of thetriplet-of-characters « »,
seven
7x4. It is also equal to the number of Arabic
alphabets. Once again this specific keyword, representing the root of the word seven, seems to
respect the 7-based structure with a repetition of 28 times, which is not only a multiple of 7, but
also equal to the number of the Arabic alphabets composing the holy book. -S- (Ref. Statistics
made in April 2015).
10. We strangely noticed that the total occurrence number of number seven« / / »,
representing the number 7 used as Noun All the Seven are present”, is
exactly 7. Herein, number seven is used seven times, but why only when it is employed as noun?
Maybe it is an indication on its profound entity/symbol and not in its function. -S- (Ref. Statistics
made in April 2015).
11. We strangely noticed that the total occurrence number of number seven « » used without prefix
and without suffix, is exactly 14= 7
number by any computer, without looking for any additional prefix or suffix. -S- (Ref. Statistics
made in April 2015).
12. We strangely noticed that the total occurrence number of the word « », meaning « Hidden-
secrets/ the Unseen/ Knowledge-on-the-future», in the holy Quran, is exactly 49 =7x7 (double
multiplicity by 7). This word, which refers to unseen things or hidden secrets, is closely related to
the subject of our main investigation and should represent a serious keyword to investigate. -S-
(Ref. Statistics made in April 2015).
13. We strangely noticed that the total occurrence number of the word « », meaning « The
Science », in the holy Quran, is exactly 28 (=4x7), a multiple of 7. Again, if we count the
(=3x5x7), which
is still a multiple of 7. It is clear that all fields of numerical analysis or Mathematics represent a
particular significant word and probably much more scientific evidences in the holy book. -S- (Ref.
Statistics made in April 2015).
14. We strangely noticed that the total occurrence number of the adjective/name (of God) « »,
meaning « Knowing everything (God)», in the holy Quran, is exactly 161 (=23x7) -S- Note also
that the Quran had been sent down for a period of 23 years (Ref. Statistics made in April 2015).
15. We strangely noticed that the total occurrence number of thecomposed word « »,
meaning « Those who possess intellect/ who reflect (on the origin of creation/ on the matters of the
Quran)», in the holy book, is exactly 7. It should probably exist a subtle relationship between this
composed word and the mathematical structure, which we are looking for: hence, the seven
repetitions of this composed word could not be in vain. -S- (Ref. Statistics made by H. Sayoud in
April 2015).
16. We strangely noticed that the total occurrence number of the expression/adjective « »,
meaning « There is no doubt in it (Quran/ Judgement day)», in the holy Quran, is exactly 14 =2x7
(multiple of 7) In the opinion of the author, this expression is one of the most important keywords
that could possess a specific mathematical code in the holy Quran, since it clearly states that
-IS-NO-DOUBT-IN- total occurrence number of
that expression is a multiple of 7, confirming the previous opinion. Moreover, we strangely noticed
that the number of alphabetical characters used in this expression is exactly 7 alphabets (
© Copyright 2021-2025 Sayoud.net. All Rights Reserved. 88
), which represents another surprising fact. Again, in the opinion of the author, this expression
has a profound significance and semantically states that there is no doubt in the Divine origin of
the holy scripture. On the other hand, it statistically shows an extreme intelligence behind those
words. -S- (Ref. Statistics made in April 2015).
17. We strangely noticed that the total occurrence number of the word « » without suffixes, meaning
84 =7x12 (multiple of 7). On the other hand, by looking for any
in the holy book, we can easily discover that the total number of verses in the holy Quran, by
keeping the bismallas (i.e. 6236+112), is equal to 6348 =12x23x23. Hence, it strangely appears that
number 12 is a common factor between both equalities, but we do not know what is the real reason
for that coincidence. We also notice, in the number of verses equation (12x23x23), that 23 could
represent the Quran revelation period, which is really 23 years indeed, and then by multiplying it
to 12 (e.g. twelve months), we also get the total number of months concerning the revelation period,
namely: 12x23 months. Sincerely, we cannot argue that this is a pure hazard.-S- (Ref. Statistics
made in April 2015).
18. We strangely noticed that the total occurrence number of the word «
» when related to the holy
book, meaning « Revelation (of the holy book)», is 14=7x2. Note that it is employed 14 times in
relation with the holy book (meaning: Revelation), and once (another time) with regards to the
Angels with another meaning (i.e. Coming down). Moreover, all the occurrences of this word (with
the two meanings) are present in exactly 14 (7x2) different chapters: 17, 20, 25, 26, 32, 36, 39, 40,
41, 45, 46, 56, 69 and chapter 76. Strangely again, the number of concerned chapters is also a
multiple of 7. -S- (Ref. Statistics made in April 2015).
19. We strangely noticed that the total occurrence number of theword « / », meaning
« Quran/ Criterion/ Salvation», in the holy Quran, is exactly 7. This word, which usually refers to
what is found, herein, proposes the probable existence of a strong numerical structure in the holy
book. -S- (Ref. Statistics made in April 2015).
20. We strangely noticed that the total occurrence number of theword « », meaning
« Justice/Equality. », in the holy Quran, is exactly 14= x2 -S- (Ref. Statistics made in April 2015).
21. We strangely noticed that the total occurrence number of the expression/adjective «
», meaning « (God) is able to do all things », in the holy Quran, is exactly (=5x7) -S- (Ref.
Statistics made in April 2015)
22. We strangely noticed that the total occurrence number of the double- «
», meaning « The Forgiving, the Merciful. », in the holy Quran, is exactly 7. -S- (Ref.
Statistics made in April 2015).
23. We strangely noticed that the total occurrence number of thedouble-word adjective « »,
meaning « (is) Forgiving, Merciful. », in the holy Quran, is exactly 49 =7x7 (double multiplicity
by 7). -S- (Ref. Statistics made in April 2015).
24. We strangely noticed that the total occurrence number of the double- «
», meaning « the Lord of the Worlds », in the holy Quran, is exactly 4 (=7x). -S- (Ref. Statistics
made in April 2015).
© Copyright 2021-2025 Sayoud.net. All Rights Reserved. 89
25. We strangely noticed that the total occurrence number of the double- «
», meaning « The Receiver of repentance, the Merciful. », in the holy Quran, is exactly 7. -
S- (Ref. Statistics made in April 2015).
26. We strangely noticed that the total occurrence number of the double- «
», meaning « the Almighty, the Wise », in the holy Quran, is exactly 4 (=7x). -S- (Ref.
Statistics made in April 2015).
27. We strangely noticed that the total occurrence number of thedouble-word adjective « »,
meaning « (Allah is) Severe in punishment », in the holy Quran, is exactly (=7x). -S- (Ref.
Statistics made in April 2015).
28. We strangely noticed that the total occurrence number of the double-word « », meaning
« (Allah gives of His Bounty) Without limit. », in the holy Quran, is exactly 7. -S- (Ref. Statistics
made in April 2015).
29. We strangely noticed that the total occurrence number of the double-word related to God: «
», meaning «(God is possessor of) the Sublime Grace », in the holy Quran, is exactly 7. -S-
(Ref. Statistics made in April 2015).
30. We strangely noticed that the total occurrence number of the double-word «
», meaning « All affairs/ Decisions are returned back (to Allah) », in the holy Quran, is exactly 7.
-S- (Ref. Statistics made in April 2015).
31. We strangely noticed that the total occurrence number of theword « », meaning « The
Believers», in the holy Quran, is exactly (=7x) -S- (Ref. Statistics made in April 2015).
32. We strangely noticed that the total occurrence number of the word « », meaning
« The Believers », in the holy Quran, is exactly (=7x) -S- (Ref. Statistics made in April 2015).
33. We strangely noticed that the total occurrence number of theword « », meaning
« Female Believer(s) », in the holy Quran, is exactly (=7x) -S- (Ref. Statistics made in April
2015).
34. We strangely noticed that the total occurrence number of the word « », meaning « the
Spirit », in the holy Quran, is exactly (=7x) -S- (Ref. Statistics made in April 2015).
35. We strangely noticed that the occurrence number of the expression « », meaning
« God is sufficient for (protecting) us/me/him », in the holy Quran, is exactly 7. -S- (Ref. Statistics
made in April 2015).
36. We strangely noticed that the total number of the 3rd Arabic pronoun «
/
», meaning «She/
He », is respectively 65 (13x5) + 481 (13x37) = 546 =7x78 (multiple of 7 too). -Note that the
pronoun “it” doesn’t exist in Arabic (it is replaced by He or She)-. Another strange relationship
between those two pronouns is the number 13 that is present in the 2 equalities. -S- (Ref. Statistics
made in April 2015).
37. We strangely noticed that the total occurrence number of the double-«
», meaning « (Allah is) all-Pervading, all-Knowing », in the holy Quran, is exactly 7. -S- (Ref.
Statistics made in April 2015).
38. A particular interest is given to the word « », which means « All Praise (is due to Allah) ». This
one is the first word in the whole text of the holy Quran after the Bismala and is often used as the
© Copyright 2021-2025 Sayoud.net. All Rights Reserved. 90
first word in several chapters too, which makes it quite important. Moreover, Muslims repeat it in
their prayers (Fatiha chapter) at least 17 times per day and the word is exclusively reserved to God.
In terms of statistics, it is cited 28 times: 28=7x4 (multiple of 7). -S- (Ref. Statistics made in April
2015).
39. We strangely noticed that the total occurrence number of theword « », meaning « good
tidings/ good news », in the holy Quran, is exactly=14=2x7. -S- (Ref. Statistics made in April
2015).
40. We strangely noticed that the total occurrence number of the word «
exactly 7. -S- (Ref. Statistics made in April 2015).
41. We strangely noticed that the total occurrence number of the word «
(i.e. future auxiliary4 =7x (multiple of 7). -S- (Ref. Statistics made in April 2015).
42. We strangely noticed that the total occurrence number of the verb «
7 (Ref. Statistics made in April 2016).
43. We strangely noticed that the total occurrence number of the word «
4 =7x (multiple of 7) (Ref. Statistics made in April 2016).
44. We strangely noticed that the total occurrence number of the word «
is exactly21 =7x3 (multiple of 7) (Ref. Statistics made in April 2016).
15.4 Discussion
From this investigation and according to the numerous results showing the presence of a 7-based multiplicity
in many parts of the holy Quran, we can state that it should exist a sort of signature or watermarking based on
holy book. Consequently, the possibility that a human being could have invented such a complex scripture
appears to be impossible. Again, the probability that the holy book could be falsified or altered is consequently
extremely weak.
Concerning the author style difference between the Quran and Hadith, this new result shows that the Quran
style that appears to embed various numerical signatures, is really different from the Hadith style, which does
not contain any such signature.
In the verse: «
» (15:9), it is clearly stated that the Holy Scripture is/will be protected
and preserved by His Creator (Ibnu-Kathir, s. d.). So, it is noticeable that the new numerical structure, which
has been discovered in this investigation, could not be a hazard at all. The above results make the hypothesis
advanced by those who think that the Quran could be a human invention illogical. How could a man (the
Prophet), who was illiterate, develop a so sophisticated mathematical structure, for preserving his scripture,
fourteen centuries ago?
© Copyright 2021-2025 Sayoud.net. All Rights Reserved. 91
16 Tenth Series of Experiments: Authorship Attribution
based on the Interrogative Form
In this section, we tackle the problem of author discrimination between the two religious books, by segmenting
the text documents into several segments. The originality of this research work lies in the use of a new set of
linguistic features based on 26 interrogative features and a special fusion. The different experiments are
performed by using a statistical measure and 8 Machine-Learning classifiers, namely: SVM, MLP, Naïve-
Bayes, Bayes Network, Simple-Logistic Regression, Voted Perceptron, J48-Decision and Random-Forest.
Acknowledgements: This research work had been made in cooperation with my PhD student Hassina Hadjadj.
16.1 Text Segmentation
A text segmentation is applied in order to construct individual documents with the same size. In fact, when
comparing two books with different sizes, it is difficult to know if a specific part of the book is similar to another
one or not. That is why a smart segmentation has been proposed and applied to the different books.
The sizes of the segments are more or less in the same range: we have 29 different text segments for the Quran
and 8 different text segments for the Hadith, with approximately the same size. Hence, we get 37 different text
segments of about 2900 words each.
The segmented dataset is decomposed into 2 parts: training part and testing part, and since the two books have
different sizes (29 texts for the Quran and 8 texts for the Hadith), a logical rule has been established: 4 text
segments are used for the training of the Hadith book and 8 text segments are employed for the training of the
Quran book. And the remaining text segments are used for the testing step.
The choice of the training dataset size is defined by a special parameter we called big/small, which gives a
qualitative estimation on the size of the book.
16.2 Proposed Interrogative Features
In our investigation, a mixture of 26 interrogative features is proposed. All those features are original and
used for the first time in stylometry (when writing this article).
Why Interrogative features?
An important step consists in retrieving distinctive features that exhibit the writing style and represent a certain
authorship individually.
to the language a unique character and specificity. However, when we read the two books, we notice that the
Interrogative form is widely used within the books. The interrogative form is one of the specificities that
characterizes the Arabic language and has a great variety and richness in the literary style.
During the preparation of this research work, there were no published works in stylometry using interrogative
features, especially in Arabic, despite the importance of the interrogative style in all languages of the world.
This fact encourages the idea of establishing a research in Arabic authorship discrimination using those
interrogative features.
Questioning features in Arabic
Each language uses morphemes and specific interrogative structures, and the interpretation of the meaning
depends on the social and cultural status of the speakers. Thus, the question differs in its interpretation and
purpose concerning the communication situation. Furthermore, the question mark in Arabic has some properties
that distinguish it from the English language and other languages.
The Arabic language, like English, has two types of questions: direct questioning and indirect questioning. The
latter does not end with a question mark and has different properties from those documented in the English
language (Mohamed-Cherif, 2007) (Naghich, 2012). The Arabic interrogation is total or partial depending on
© Copyright 2021-2025 Sayoud.net. All Rights Reserved. 92
its scope, and according to its value, it can be seen as in the case of the English language: rhetoric, inquiries,
disguised order, etc.
Interrogative modality in Arabic has some characteristics that depend not only on the methods of the
interrogation, i.e. intonation, interrogative morpheme and phrasal structure, but may also vary in terms of goals
and sometimes the way of speaking.
Types of questions in Arabic
Direct and indirect inquiry
The Arabic language has two types of questions: the first is direct and it ends with a question mark, the
second is called indirect and ends with a full-stop: it is a question where the verb introducer is semantically
interrogative. Arabic, in questioning, is generally defined by two structures: one is the enquiry that is to say the
interrogative element that appears on the original sentence or "Ism Al-istifham", "harf Al-istifham" (see below)
and the other one is ensured by some specific verbs.
Do not you see he is a foreigner?
I wonder if you see that he is a foreigner
Question 1 represents a direct query. However, it can be ensured by using some specific verbs such as the verb
"TasâAla " (meaning: I wonder if) to become an indirect question, such as in question 2. The verb "TasâAla
" can be prefixed with the morphological tag bearing the features of the type of the subject in the unfinished
appearance (a: I), (ta: you ), (there: he, she, they) and (na: us).The indirect question is introduced by a verb
like "TasâAla ", for instance
The total and partial query
Depending on the scope of the question, the question is answered by an affirmation or a reversal of the
content of the statement. In this case, it is called total interrogation, and is introduced by the interrogative
element, namely "Harf Al-istifham?" ("Hal " and "Hamza aa ", which are both equivalent to the auxiliary
verb do in English) as shown in the following example below. Both sentences have the same meaning and the
expected answer is yes or no.
Did Ahmed come?
Has Ahmed entered?
The total indirect and polling can be introduced by a verb in interrogative semantic such as "tasâAla " or
interrogative proposal such as "uridu an aårifa: " as shown in the following examples :
I want to know if Ahmed came.
I wonder if Ahmed entered.
The partial query is introduced by an interrogative pronoun called "Ism Al-istifhâm" such as ( : what, :
who, : how, : where). It can be, like the total query, direct or indirect. In the case of a partial direct
question, the interrogative pronoun is at the initial position of the interrogative sentence. In the case of a total
© Copyright 2021-2025 Sayoud.net. All Rights Reserved. 93
indirect question, the interrogative pronoun is replaced by a subordinating conjunction that introduces the
sentence, as shown in the following examples:
Where was you gone?
I wonder where you went
The semantic features of interrogative prepositions
The interrogative modality in English may be denoted through the ascending intonation or the presence of
an interrogative element to the initial sentence. Another method is the indirect question, which is marked by the
subordinating conjunction (if) or (that). However, in Arabic, there are several other conjunctions, for example:
inn (if), idâ (if), ma idâ (in case of).
Interrogative morphemes are used to check the trueness of a statement or question in the interrogative sentence.
Pronouns
• / --
trueness of a fact, for example:
- Did he come ?
• --
- Did he come ?
• /
act, for example:
- What are you doing?
• / "/ mâdâ, for example:
- What made youunable to come today?
- What do you have?
• /
- Which guest came?
• / concerned with the subject or a
determinative pronoun.
- Who says that?
Adverb
When the question is submitted by an interrogative adverb, it becomes an adverbial phrase.
• / lima"): it is used to request the cause of an event.
There is also the variation"/ bimâdâ" meaning "with what", which is used to build an adverbial question,
example:
- Why do you come now?
- With what you write?
•
© Copyright 2021-2025 Sayoud.net. All Rights Reserved. 94
- How did you get here?
- How is your new car?
- How are you?
•
quantity, example:
- How many days you were absent?
• / questions about the time.
- When will you come back?
•
The particle "Am" is used concurrently with the interrogative particle "Aa", which is sometimes implied.
This particle is used when the speaker knows that the fact is true, and wonders who/which associated
person/thing is concerned with, example:
Who is with you: Zayd or Amr? /
Interrogative Features selection
According to the importance of the interrogative style in our investigation, a set of 26 interrogative features was
selected and extracted and collected from the two religious books.
This set contains four types of interrogative features as follows:
• Interrogative words/verbs with the particle Hamza ( / aa): this type of interrogation is commonly used
form. In our approach, we employed 22 interrogative elements of this type.
•
questioning.
• The starting particle
sentence, it refers to a form of question.
•
interesting since it gives a feeling of a strong personality when speaking.
•
interrogative expression is also interesting for its strong style when speaking.
16.3 Authorship Discrimination Approach
In our method, several steps are performed, as shown in Figure 16.1, namely: data pre-processing, feature
extraction, classification and author discrimination decision, while the data set is organized into training and
testing.
© Copyright 2021-2025 Sayoud.net. All Rights Reserved. 95
Figure16.1. Authorship discrimination method
16.4 Feature extraction and LFF fusion
This is an important step in author identification. In our case, we recall that the originality of this investigation
consists in the use of the proposed interrogative features.
Furthermore, we have defined the Logarithmic Feature Fusion (LFF) as the logarithmic sum of the normalised
features frequencies (see equation 16.1). A graphical representation of the LFFs for the 37 text segments is
displayed in figure 16.2.
We can notice that all the Quran LFFs are positive while the Hadith LFFs are negative.
The Logarithmic Feature Fusion (LFF) is given as follows:
whith i representing the feature index (i=1,2,...26).
Figure 16.2 shows a sharp difference between the Quran segments, which present positive logarithmic
values, and the Hadith segments, for which the logarithmic values are negative.
Testing
data
Interrogative features
Extraction
Data preprocessing
Interrogative features
Extraction
Reference Authors
Models
Training
data
Model Classification
Most likely model
(Author)
Data preprocessing
Unknown Author
Model
© Copyright 2021-2025 Sayoud.net. All Rights Reserved. 96
Figure.16.2. LFF values of the Quran text segments (in dark green) and Hadith text segments (in pink)
16.5 Classification methods
Description of the classifiers
The different experiments of authorship attribution are made by using several classifiers, namely:
• Euclidean Distance
• Multi-layer-Perceptron
• Naïve-Bayes classifier
• Bayesian-Network
• Support Vector Machines
• Voted Perceptron
• J48-Decision-Trees
• Random-Forest
• Simple-Logistic-Regression
Performance evaluation
The performance given by the classifier is measured in terms of accuracy of classification, which is
calculated as the ratio of the number of correct attributions over the total number of testing segments.
(16.2)
16.6 Experimental results and analysis
This investigation performs a segmental analysis on the two religious books: Quran and Bukhari
Hadith, for the task of authorship discrimination. The analysis of the different text segments is performed
-2.0
-1.5
-1.0
-0.5
0.0
0.5
1.0
Q1
Q2
Q3
Q4
Q5
Q6
Q7
Q8
Q9
Q10
Q11
Q12
Q13
Q14
Q15
Q16
Q17
Q18
Q19
Q20
Q21
Q22
Q23
Q24
Q25
Q26
Q27
Q28
Q29
H1
H2
H3
H4
H5
H6
H7
H8
LFF
Q1 Q2 Q3 Q4 Q5 Q6 Q7 Q8 Q9 Q10 Q11 Q12 Q13
Q14 Q15 Q16 Q17 Q18 Q19 Q20 Q21 Q22 Q23 Q24 Q25 Q26
Q27 Q28 Q29 H1 H2 H3 H4 H5 H6 H7 H8
© Copyright 2021-2025 Sayoud.net. All Rights Reserved. 97
by using 27 interrogative features (i.e. 26 interrogative features + their fusion) and by employing 9
different classifiers.
The dataset is composed of 37 different text segments of about 2900 words each, consisting of 8
the experiments,
4 segments of the Hadith and 8 segments of the Quran are used for the training and the remaining
segments are used for the testing. Therefore, there are 25 different segments to classify according to 2
referential Authors (Quran Author or Hadith Author).
Table 16.1 shows the A.A. accuracy on the two religious books. From this table, we notice that the
proposed feature set is powerful for discriminating the two books styles and we can notice that all the
at
say that the new proposed features have efficiently separated the two books according to their
interrogative styles.
So, the proposed feature set appears to be effective in authorship attribution, especially with the
Arabic language. Furthermore, since only interrogative features are employed, this analysis represents
an evaluation of the interrogative style in each book.
Experimentally speaking, we noticed that the interrogative styles in the Quran and Bukhari Hadith
are completely different (an accuracy of 100% is achieved with all the used classifiers).
TABLE 16.1. Authorship Attribution accuracy on the 2 religious books by using the new proposed
features
Classification
Algorithm
Accuracy
Attribution Error
Euclidean Distance
100%
0%
SMO-SVM
100%
0%
MLP
100%
0%
Naïve-Bayes
100%
0%
Bayes-Network
100%
0%
Simple-Logistic-
Regression
100%
0%
Voted Perceptron
100%
0%
J48-decision Tree
100%
0%
Random-Forest
100%
0%
Notes: It has been noticed that the particle hamza is very commonly used in the Quran book; whereas,
in the Hadith, this particle is rarely used. Again, we noticed that the following interrogative expressions
« » are more commonly used in the Quran; but almost never used in the Hadith.
Moreover, we also noticed that the Author of the Quran used indirect questioning form quite frequently,
whereas in the Hadith it is not used at all, which shows that the two authors possess two distinct
interrogative styles that are completely different.
According to the structure and the morphology of the interrogative style in the Arabic language, it is
very difficult to write two books in the same topic with so many differences in the style, and we do not
see any other explanation except the fact that the Quran and Hadith should have two different authors.
© Copyright 2021-2025 Sayoud.net. All Rights Reserved. 98
16.7 Discussion
This research work has addressed the problem of author classification by segmenting the concerned
documents into several texts segments of the same size. The originality of this research work lies in the
use of a new set of features based on 26 interrogative features exclusively and a special fusion; and it is
the first time that those features are used alone in stylometry.
That is, different classification algorithms were used to build an automatic classification model to
discriminate between the authors: Centroid-Euclidean distance, Support Vector Machines, Multi-Layer-
Perceptron, Naïve-bayes, Bayes-Network, Simple-Logistic-Regression, Voted-Perceptron, J48-
Decision-tree algorithm and Random-Forest.
A specific application of this research work concerned the problem of author discrimination between
two old Arabic religious books: The Quran and Hadith to check whether the two books could have been
written by the same author or not.
Basically, the different experiments of authorship attribution have led to the following important points:
▪ Concerning the proposed interrogative features, they appear to be interesting for performing
author identification tasks in Arabic;
▪ Concerning our application of author discrimination between the two religious books, the
experimental results have clearly revealed that the interrogative styles structures of the Quran and
Hadith are very different.
Now, despite this result, if we assume that the two books belong to the same Author, then an
important arising question would be: Could an author possess two interrogative styles completely
different for the same topic?
If so, what could be the reason to change his questioning style? And in what proportions would it be
possible to do that variability?
The only possible response to all those questions leads to the fact that there could be no valid reason
to use two dissimilar interrogative styles completely different by the same author.
Actually, we do not see any other explanation except the fact that the two studied books should have
two different Authors.
© Copyright 2021-2025 Sayoud.net. All Rights Reserved. 99
17 Eleventh Series of Experiments: Investigation on the
Quran/Hadith Authorship Using Visual Analytics
Approaches
An important raising question, in the stylometric study of the Quran, is: Was this religious book written by
the Prophet? An interesting scientific way, to answer that question, is the use of authorship attribution
techniques. However the use of conventional features and classifiers has some disadvantages such as the
automatic authorship decision, which usually gives us a speechless authorship classification without (often)
any way to measure or interpret the consistency of the results.
In this research work, we present a visual analytics based investigation for the task of authorship
discrimination between the holy Quran and Hadith.
Seven types of features are combined and normalized by PCA (Principal Components Analysis) reduction
and seven visual analytics based clustering methods are employed and commented, namely: Hierarchical
Clustering, Fuzzy C-mean Clustering, K-mean Clustering, Sammon Mapping, Principal Component
Analysis, Gaussian Mixtures Models and Self Organizing Maps.
Results are quite interesting and the disposition of the visual clusters provide valuable information on the
.
17.1 Stylometric features
Seven types of features have been proposed for this task of AA, as one can see in the following subsections.
17.1.1. Author’s Pronoun Based Stylistics
) and We ( - ) are the most used ones for
representing Allah (God) in the holy Quran. See some examples below:
- (He) design to the heavens and Hisit is who has created for you all that is on earth, and has applied He“
alone has full knowledge of everything” Hefashioned them into seven heavens; and
- (We)
appointed for Mûsa (Moses) forty nights, and (in his absence) you took We“And (remember) when
the calf (for worship), and you were Zâlimûn (polytheists and wrong-doers)”
However, in some few cases, the pronoun I ( - ) is also used, but in special circumstances. See some
examples below: - ( I / Me) " Me, and do not deny Meshall remember you; and be grateful unto I, and Me“So remember -( I / Me ) respond to the Iam near. Iindeed – Meservants ask you, (O Muhammad) ,concerning My“And when (by obedience) and Me. So let them respond to Meinvocation of the supplicant when he calls upon
so that they may be (rightly) guided" Mebelieve in
Perhaps, in those last examples, the pronoun I is used for representing the closeness of Allah (God) to
his believers, or perhaps, the pronoun I is used for meaning that Allah (God) is ONE, but when speaking
about God statements, Allah We
© Copyright 2021-2025 Sayoud.net. All Rights Reserved. 100
Consequently, in the Quran, we notice that Allah (God) uses the 3 pronouns: Me / I (- ), We (-
) and He () for representing His Excellency Allah.
Differently, in the Hadith, the Prophet Muhammad uses exclusively the pronoun: Me / I (- ) for
representing himself.
17.1.2. On the use of “
“ (father of) for naming people
In the Arabic language, it is usual to call a person using the name of his oldest child (often the son). That
is, if somebody has a son called Youssof for instance, then it is possible to call him Aba-Youssof, which
can be translated in English into Father-of-Youssof. This fact was often noticed in the Hadith, when the
Prophet speaks to his companions.
Nowadays, although this mean of appellation has become rare, it is still widely employed in some
countries of the Arabic gulf region and middle-east.
On the other hand, namely in the Quran, the style of appellation is quite different, since the persons are
directly called by their own names (first-names), such as in the following verse:
(78)
Which could be translated into: And [mention] David and Solomon, when they judged concerning the
field - when the sheep of a people overran it [at night], and We were witness to their judgement ( 78 ).
17.1.3. Frequency of some discriminative words
As first feature, we have proposed to use some words that are very commonly used in only one of the
books. In practice, we remarked that the words: (in English: THOSE or WHO in a plural form) and
(in English: EARTH) are very commonly used in the Quran book; whereas, in the Hadith, these
words are rarely used.
17.1.4. On the COST parameter
Usually, when poets write a series of poems, they make a termination similarity between the neighboring
sentences of the poem, such as a same final syllable or letter. To evaluate that termination similarity, a
new parameter estimating the degree of text chain (in a text of several sentences) has been proposed: the
COST parameter (H. Sayoud, 2012). The description of this parameter can be found in section 7.3
17.1.5. Word length frequency
The fifth feature is the word length frequency. Herein, the word length is the number of letters
composing that word and the word length frequency F(n) for a
(in percent) of words composed of n letters each, present in the text.
17.1.6. Frequency of the coordination conjunction «
» (meaning AND in English)
An interesting type of features corresponds to the coordination conjunctions, which are widely used in
the two investigated books.
© Copyright 2021-2025 Sayoud.net. All Rights Reserved. 101
In reality, we limited our investigation to one of the most interesting conjunction, which seems to be
coordination conjunction AND (in English) and which is widely used in Arabic.
17.1.7. Frequency of the coordination conjunction Waw «
» at the beginning of
sentence
Similarly to the previous section, herein we are still interested in the frequency of the coordination
And
In fact, we noticed that several verses in the Quran begin with that coordination conjunction, which is
because as its appellation says (i.e. coordination), it connects/coordinates two sentences, two verbs or
two names and consequently it may not be localized at the beginning of a sentence, except in few rare
cases.
17.2 Visual Analytics based Clustering methods
By definition, the term clustering corresponds to the fact of grouping some things together; which can
be physical objects, numerical data, concepts or any sort of elements.
In pattern recognition, cluster analysis or clustering is the task of grouping a set of objects in such a way
that objects in the same group (ie. cluster) are more similar to each other than to those in other groups
(Norusis, 2008). This task is commonly used in data mining, statistical data analysis, machine learning
and information retrieval.
On the other hand, visual Analytics (Ellis et al., 2010), which is a combination of several fields (ie.
computer science, information visualization and graphic design) is often used in cluster analysis to make
That is, the combination of those two research fields can lead to a strong and efficient analysis tool for
handling some classification tasks that could be extremely difficult to perform with conventional
analytic tools.
Furthermore, a great advantage of clustering over conventional classification tools is its non-supervised
property (for several clustering techniques).
Consequently, it appears that the association of visual analytics with clustering analysis may be
interesting for solving some stylometric problems, for which we do not possess any training possibility
or information to make a supervised classification task. So, it should be extremely motivating to apply
them in our main task of authorship discrimination (ie. Quran vs Hadith).
Concerning the methods using the association of visual analytics with clustering analysis, there exist
several approaches that have been proposed during the last five decades, such as: K-mean Clustering,
Hierarchical Clustering, Sammon Mapping, Self Organizing Maps, Gaussian Mixtures Models, Fuzzy
C-mean Clustering, Principal Component Analysis, etc.
In this survey, we propose to use all those seven methods separately in order to find out the possible
clusters related to the different investigated texts.
© Copyright 2021-2025 Sayoud.net. All Rights Reserved. 102
17.2.1 Hierarchical clustering
Our first clustering method is based on the hierarchical clustering.
Definition
Hierarchical clustering is a method of cluster analysis which seeks to build a hierarchy of clusters
(Greenacre, 2014). In general, there are two types:
-Agglomerative clustering: This is a "bottom up" approach, where each observation starts in its own
cluster, and pairs of clusters are merged as one moves up the hierarchy.
-Divisive clustering: This is a "top down" approach, where all observations start in one cluster, and splits
are performed recursively as one moves down the hierarchy.
In our case, we used the first clustering type with a Manhattan distance measure, which is defined below:
If we assume that X and Y represent two vectors, then Manhattan distance (between those 2 vectors) is
given by the following equation:
(1)
The resulting linkage of the
possible clusters in a graphical way. By observing the dendrogram, it will be possible to estimate the
actual number of clusters and the corresponding documents for each cluster, since all similar documents
should be linked together with a consistent linkage.
Results of the Hierarchical clustering
The hierarchical clustering has yield to the following dendrogram, where we can observe two separate
clusters, one cluster in red in the right and another one in blue in the left. The corresponding result shows
clearly that the two investigated documents (i.e. Quran and Hadith) have two different styles.
Figure 17.1: Results of the Hierarchical Clustering
H11 H02 H06 H05 H07 H03 H08 H0 1 H09 H10 H04 Q01 Q04 Q09 Q06 Q10 Q13 Q14 Q07 Q11 Q12 Q05 Q08 Q02 Q03
0
1
2
3
4
5
Text segment
Distance
© Copyright 2021-2025 Sayoud.net. All Rights Reserved. 103
17.2.2 C-Means clustering
Definition
Fuzzy clustering is a class of algorithms for cluster analysis in which the allocation of data points to
clusters is not "hard" (all or nothing) but "fuzzy" in the same sense as fuzzy logic (Suganya, 2012).
In fuzzy clustering, every point has a degree of belonging to clusters, rather than belonging completely
to just one cluster. Thus, points on the edge of a cluster, may be in the cluster to a lesser degree than
points in the center of cluster.
That is, any point x has a set of coefficients giving the degree of being in the kth cluster wk(x). With fuzzy
c-means, the centroid of a cluster is the mean of all points, weighted by their degree of belonging to the
cluster:
(2)
In 3D or 2D dimensions, Fuzzy C-mean can provide a graphical representation of the different samples
and the corresponding clusters to which they should belong. This representation allows separating the
different samples with regards to their similarities automatically and in a visual manner.
Results of the Fuzzy C-Means clustering
The Fuzzy C-Means clustering has provided the following 3D representation, where we can observe
two separate clusters, one cluster in red in the right and another one in blue in the left. This result shows
that the two investigated documents (i.e. Quran and Hadith) should have two different styles.
© Copyright 2021-2025 Sayoud.net. All Rights Reserved. 104
Figure 17.2: Fuzzy FCM Clustering
17.2.3 K-Means clustering
Definition
K-means clustering is a method of vector quantization, originally from signal processing, that is popular
for cluster analysis (Kardi, 2007).
K-means clustering aims to partition n observations into k clusters in which each observation belongs
to the cluster with the nearest mean, serving as a prototype of the cluster.
So, given a set of observations (x1, x2xn), where each observation is a d-dimensional real vector, k-
means clustering aims to partition the n observations into k n) sets S = {S1, S2Sk} so as to minimize
the within-cluster sum of squares (WCSS). In other words, its main objective is to find the following
value:
²
(3)
00.2 0.4 0.6 0.8 1
0.16
0.18
0.2
0.22
0.24
0.5
0.55
0.6
0.65
0.7
0.75
0.8
0.85
0.9
0.95
1
Feature x
Feature y
Feature z
Quran Cluster
Hadith Cluster
© Copyright 2021-2025 Sayoud.net. All Rights Reserved. 105
where μi is the mean of points in Si.
As in Fuzzy C-mean, the K-means can provide a graphical representation of the different samples and
the corresponding clusters to which they should belong. This representation allows separating the
different samples with regards to their similarities in an automatic way.
Results of the K-Means clustering
The K-means based clustering led to the following 3D representation, where we can easily notice that
the different text segments have been grouped into two main clusters: a Quran cluster in the right and
Hadith cluster in the left of the 3D representation. This sharp separation suggests that the two types of
texts should have two different authors.
Figure 17.3: K-means Clustering.
17.2.4 Sammon Mapping
Definition
Sammon mapping or Sammon projection is an algorithm that maps a high-dimensional space to a space
of lower dimensionality by trying to preserve the structure of inter-point distances in high-dimensional
space in the lower-dimension projection (Kim, 2003). It is particularly suited for use in exploratory data
00.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
0.16
0.18
0.2
0.22
0.24
0.5
0.55
0.6
0.65
0.7
0.75
0.8
0.85
0.9
0.95
1
Feature x
Kmean Clustering
Feature y
Feature z
Quran Cluster
Hadith Cluster
© Copyright 2021-2025 Sayoud.net. All Rights Reserved. 106
analysis. The method was proposed by John W. Sammon in 1969. It is considered a non-linear approach
since the mapping cannot be represented as a linear combination of the original variables such as in the
principal component analysis.
That is, by denoting the distance between the ith and jth elements in the original space by d*ij, and the
distance between their projections by dij. Sammon's mapping aims to minimize the following error
function, which is often called Sammon's error:
²
(4)
By choosing a 3 or 2 dimensional space, the Sammon-based graphical representation is quite interesting,
since it makes a sharp separation of the different elements by bringing closer all the similar ones.
Results of the Sammon mapping
A 3 dimensional reduction has been employed using Sammon representation. The 3 retained features
are denoted by alpha, beta and gamma. Thereafter, the 25 text samples are represented according to
those 3D axes/features. The resulting visual representation shows 2 main clusters: one in the right
grouping all the Hadith segments and another one in the left grouping all the Quran segments.
Furthermore those two sets of texts are covered by an interpolated surface between samples of a same
type in order to better show the different resulting clusters. In other words, those surfaces are not
provided by Sammon mapping but by our personal algorithm for a visual confort only.
© Copyright 2021-2025 Sayoud.net. All Rights Reserved. 107
Figure 17.4: Sammon Mapping and intra-elements surface interpolation
17.2.5 Principal Components Analysis
Definition
Principal component analysis (PCA) can be considered as one of the most interesting results of applied
linear algebra. PCA is used abundantly in all forms of analysis - from neuroscience to computer graphics,
because it is a simple and non-parametric method of extracting relevant information from confusing data
sets. With minimal additional effort PCA provides a roadmap for how to reduce a complex data set to a
lower dimension to reveal the sometimes hidden, simplified dynamics that often underlie it (Shlens,
2003).
PCA is mathematically defined as an orthogonal linear transformation that transforms the data to a new
coordinate system such that the greatest variance by some projection of the data comes to lie on the first
coordinate (i.e. the first principal component), the second greatest variance on the second coordinate,
and so on.
Consider a data matrix, X, where the sample mean of each column has been shifted to zero and where
each of the n rows represents a different repetition of the experiment, and each of the p columns gives a
particular kind of datum (e.g. the results from a particular sensor). Mathematically, the transformation
is defined by a set of p-dimensional vectors of weights or loadings W(k)=(w1p)(k) that map each row
vector X(i) of X to a new vector of principal component scores t(i)=(t1p)(i) , given by
tk(i)= X(i) . W(k)
(5)
in such a way that the individual variables of t considered over the data set successively inherit the
maximum possible variance from x, with each loading vector w constrained to be a unit vector.
-3
-2.5
-2
-1.5
-1
-0.5
-1.5
-1
-0.5
0
0.5
1
-1
-0.5
0
0.5
feature alpha
Sammon mapping
feature beta
feature gamma
Quran
Hadith
© Copyright 2021-2025 Sayoud.net. All Rights Reserved. 108
PCA is quite interesting in complex data analysis, when the most important features are not known in
advance. And by reducing the dimensionality to a lower more consistent one, the data analysis becomes
usually easier and more pertinent.
Results of the PCA analysis
A PCA representation of the data, using the 3 most important eigenvectors, is given in the following
figure. The Quran texts are symbolized by red circles, while the Hadith texts are symbolized by blue
crosses. In that figure, we can notice that all the Quran documents are grouped together in the right side,
while all the Hadith ones are grouped in the left side. Once again the stylistic discrimination can be
easily noticeable in the 3D representation.
Figure 17.5: PCA representation of the data using the 3 most important eigenvectors. Quran texts are
represented by red circles while Hadith texts are represented by blue crosses.
17.2.6 Gaussian Mixture Model based clustering
Definition
Finite mixtures of distributions have provided a mathematical-based approach to the statistical
modelling of a wide variety of random phenomena (McLachlan, 2003). Because of their usefulness as
an extremely exible method of modelling, nite mixture models have continued to receive increasing
attention over the years, both from a practical and theoretical point of view. For multivariate data of a
continuous nature, attention has focused on the use of multi-variate normal components because of their
wide applicability and computational convenience. They can be easily tted iteratively by maximum
likelihood via the expectation maximization algorithm.
-1.5 -1 -0.5 00.5 11.5
-0.8
-0.6
-0.4
-0.2
0
0.2
-1
-0.8
-0.6
-0.4
-0.2
0
0.2
0.4
feature 1
PCA Analysis
feature 2
feature 3
Quran
Hadith
© Copyright 2021-2025 Sayoud.net. All Rights Reserved. 109
With a normal mixture model-based approach, it is assumed that the data to be clustered are from a
mixture of an initially specied number g of multivariate normal densities in some unknown proportions
pi1 g, That is, each data point is taken to be at realization of the mixture probability density
function.
(6)
where denotes the p-variate normal density probability function with mean , and
covariance .
Here the vector of unknown parameters consists of the mixing proportions , the elements of the
component means and the distinct elements of the component- covariance matrices .
Once the mixture model has been fitted, a probabilistic clustering of the data into g clusters can be
obtained in terms of the fitted posterior probabilities of component membership for the data. An outright
assignment of the data into g clusters is achieved by assigning each data point to the component to which
it has the highest estimated posterior probability of belonging (McLachlan, 2001).
Results of the GMM based clustering
The GMM based clustering is performed after PCA reduction into the 2 most important components.
That is, two types of visualizations are provided: a 2D representation (with those two components) and
a 3D representation including the probability density function as third component (see figures 17.6.a
and 17.6.b).
In both figures, we notice that the different text samples have been clustered into 2 main groups: Quran
cluster, at the bottom left side, gathering all the Quran texts and a Hadith cluster at top right, gathering
all Hadith texts.
In the 2D representation, the Gaussian mixtures are represented by different ellipsoids surrounding the
two clusters, while in the 3D representation, the Gaussians are more visible since they are represented
in form of 3D Gaussians surrounding the different clusters.
While, the first representation is sharper, the two representations are similar in terms of clustering
information: so, we easily notice that all Quran texts are closely clustered together and all Hadith ones
are closely grouped together too. This fact confirms, once again, that the two writing styles of the 2
books are probably different.
© Copyright 2021-2025 Sayoud.net. All Rights Reserved. 110
Figure 17.6.a: GMM clustering in 2D representation using two components.
Figure 17.6.b: GMM clustering in 3D representation. The 3rd dimension represents the probability
density function.
-3 -2.5 -2 -1.5 -1
-1
-0.5
0
0.5
1
1.5
2
1st Component
2nd Component
Clustering by GMM after PCA reduction
Cluster 1: Quran
Cluster 2: Hadith
-3 -2.5 -2 -1.5 -1
-1
-0.5
0
0.5
1
1.5
2
0
0.2
0.4
0.6
0.8
1st Component
Clustering by GMM after PCA reduction
2nd Component
pdf
Cluster 1: Quran
Cluster 2: Hadith
Quran cluster
Hadith cluster
© Copyright 2021-2025 Sayoud.net. All Rights Reserved. 111
17.2.7 Self-Organizing Map based clustering
Definition
A Self-organizing Map is a data visualization technique developed by Teuvo Kohonen in the early
1980's (Kohonen, 1990) (Tambouratzis et al., 2003).
SOMs map multidimensional data onto lower dimensional subspaces where geometric relationships
between points indicate their similarity. SOMs generate subspaces with an unsupervised learning neural
network trained with a competitive learning algorithm.
The SOM learning tries to make the different parts of the network respond similarly to certain input
patterns. This is partly motivated by how visual, auditory or other sensory information is handled in
separate parts of the cerebral cortex in the human brain.
The weights of the neurons are initialized either to small random values or sampled evenly from the
subspace spanned by the two largest principal component eigenvectors. The network must be fed a large
number of example vectors that represent, as close as possible, the kinds of vectors expected during
mapping. The examples are usually administered several times as iterations.
The training utilizes competitive learning. When a training example is fed to the network, its Euclidean
distance to all weight vectors is computed. The neuron whose weight vector is most similar to the input
is called the best matching unit. The weights of the best matching unit and neurons close to it in the
SOM lattice are adjusted towards the input vector. The magnitude of the change decreases with time
and with distance from the best matching unit. The update formula for a neuron v with weight vector
Wv(s) is
(7)
(s) is a monotonically decreasing learning coefficient and D(t) is the input vector; (u, v, s) is the
neighborhood function which gives the distance between the neuron u and the neuron v in step s.
Depending on the implementations, t can scan the training data set systematically (t = 0, 1, 2...T-1, then
repeat, T being the training sample's size), be randomly drawn from the data set, or implement some
other sampling method such as jackknifing.
The main advantage of using a SOM is that the data is easily interpreted and understood. The reduction
of dimensionality and grid clustering makes it easy to observe similarities in the data. The major
disadvantage of a SOM is that it requires necessary and sufficient data in order to develop meaningful
clusters. Moreover, lack of data will usually add randomness to the groupings.
Results of the SOMs clustering
According to the figures obtained with SOM clustering, we can easily notice that there are mainly 2
clusters. Hence, in the 3D figure below (figure 17.7.a), representing the Distance matrix (inter-
distances), We can see that there are 2 distinct dark regions (representing low inter-distances). Those 2
dark regions involve the presence of 2 distinct clusters, since every black area represents a cluster (in
this case).
In the 2D figure below (figure 17.7.b), a Self-Organizing Map (SOM) using 3 PCA components has
been performed. Here, the U-matrix is shown on the left, and an empty grid named 'Labels' is shown on
the right.
In the left figure (U-matrix), we note 2 main clusters in white (the light colors represent clusters). The
black (or dark) cells represent boundaries between clusters, unlike in the previous figure. Hence, one
big cluster is visible at the right bottom and another big one at the left top.
© Copyright 2021-2025 Sayoud.net. All Rights Reserved. 112
In the middle figure, the different cells have been labeled (with regards to the book origin) by using 2
colors (red for the Quran and green for the Hadith), showing what texts belong to each cluster.
Furthermore the distribution of the data set has been added to the map by using the corresponding hit
histograms.
In fact an important tool in data analysis using SOM is called hit histogram. It is formed by taking a data
set, finding the BMU (Best Matching Unit) of each data sample from the map, and increasing a counter
in a map unit each time it is the BMU (Best Matching Unit). The hit histogram shows the distribution
of the data set on the map. Here, the hit histogram for the whole data set is calculated and visualized on
the U-matrix.
Once again, we notice that the Quran samples in red are well grouped together and separated from the
Hadith samples in green, by a sharp horizontal black (dark) line.
Consequently, we can see that the SOM clustering leads to the same conclusion as previously, which is:
the two books should have two different authors (or at least two different writing styles).
Figure 17.7.a: 3D representation of the Distance matrix (inter-distances). We can see that there are 2
distinct dark regions (representing low inter-distances). Those 2 dark regions involve the presence of 2
distinct clusters. In other words every black area represents a cluster.
© Copyright 2021-2025 Sayoud.net. All Rights Reserved. 113
Figure 17.7.b: 2D Self-Organizing Map (SOM) using 3 PCA components. In the left figure, we can
see 2 main clusters in white. In this representation, the light colors represent clusters. Hence, one
cluster is visible at the right bottom and another one at the left top. In the middle figure, the different
cells have been labeled by using 2 colors (green for the Hadith and red for the Quran), showing what
are the cells belonging to each cluster. In the right, we only have the labels of the different SOM cells.
17.3 Results summarization
In this investigation, seven visual analytics based clustering approaches have been employed to make a
visual authorship clustering of 25 religious text segments.
The different approaches are as follows: Hierarchical Clustering (HIC), Fuzzy C-mean Clustering
(FCM), K-mean Clustering (KMC), Sammon Mapping (SAM), Principal Component Analysis (PCA),
Gaussian Mixtures Models (GMM) and Self Organizing Maps (SOM).
• In the first approach, namely HIC, the resulting dendrogram has shown two separated clusters:
the Quran cluster in the right and Hadith cluster in the left (see figure 17.1). We can also see
that there is no intersection between the different clusters and that the final linkage is extremely
weak since the corresponding distance is relatively very large. This result shows that there are
two different writing styles and then probably two different authors too: Quran Author and
Hadith Author.
• In the second approach (i.e. FCM), which is an automatic clustering technique, the resulting 3D
representation shows two main clusters: Quran cluster located at the right top area and Hadith
cluster located at the left bottom area of the 3D representation (see figure 17.2). Although the
Quran cluster is more condensed, the two sets of text segments have been automatically
organized into 2 sharp clusters (with different symbol markers), showing that there are probably
two main authors: Quran Author and Hadith Author and that the two authors are different.
• In the third approach (i.e. KMC), which is also an automatic clustering technique, the three-
dimensional K-mean representation reveals two main clusters too: Quran cluster in the right and
Hadith cluster in the left (see figure 17.3). Those two clusters are distant and well separated one
from the other. Consequently and as previously, it is clear that the two authors (Quran Author
and Hadith author) should be different.
• The fourth approach, called SAM, is not a real clustering method but only a mapping technique.
However, the resulting mapping has provided an interesting 3D representation of the texts
© Copyright 2021-2025 Sayoud.net. All Rights Reserved. 114
samples, showing two separated writing styles and consequently two different authors. This
result is easily observable in figure 17.4, where all Quran texts have been concentrated in the
left region and all Hadith text have been concentrated in the right side.
• The fifth approach (i.e. PCA) is not a clustering method either, but it is only used for feature
reduction and low dimensional mapping. The resulting 3D representation, in figure 17.5, shows
the position of the text samples thanks to their 3 first PCA coordinates. We can observe that the
Quran texts are located in the left, whereas the Hadith ones are located in the right and that the
two text groups are quite separated. We also notice that Hadith samples are more condensed
than Quran ones. The PCA representation suggests that the two books have two different author
styles.
• The sixth approach, namely GMM, is a clustering technique based on mixture models. In our
case it is used with only 2 components (2D representation). As we can see in figures 17.6.a and
17.6.b, two main GMM-based clusters have been obtained: Quran cluster at the left bottom
(grouping all Quran segments) and Hadith cluster at the right top (grouping all Hadith
segments). Consequently, and thanks to this 2D representation, the two books appear to belong
to two different authors, or at least two different writing styles.
• The seventh approach (i.e. SOM) is a self organizing neural network, which makes a 2D
representation of the different possible clusters in an interesting way, since it gives a rich amount
of information regarding the clusters and their consistency.
The resulting SOM mapping (figures 17.7.a and 17.7.b) shows a dark horizontal region
separating the different SOM cells into two main regions: Top and Bottom. The bottom area
contains Quran segments, while the top area contains Hadith segments. Furthermore, one can
observe that Hadith is represented by a big sub-cluster in the left and a small one in the right
(top area), whereas Quran is represented by a big sub-cluster in the right and a small one in the
left (bottom area). However, in the overall, the two main clusters are well separated one from
the other: Quran samples in the bottom and Hadith samples in the top, which implies that there
are 2 different author styles: one author style common to all Quran texts and another author
style common to all Hadith texts.
17.4 Discussion
The principal purpose of this investigation is to conduct some experiments of authorship discrimination
concerning two religious books: the holy Quran and Hadith, in a visual analytics way. And as described
in the beginning of this manuscript, seven different features are used to make a stylometric comparison
between the two books: Author Related Pronouns (ARP), Father Based Surname (FBS), Discriminative
Words (DisW), COST value, Word Length Frequency (WLF), Coordination Conjunction (CC) and
Starting Coordination conjunction (SCC).
On the other hand, the task of comparison is ensured by seven clustering approaches based on visual
analytics techniques. We recall those seven clustering approaches: Hierarchical Clustering (HIC), Fuzzy
C-mean Clustering (FCM), K-mean Clustering (KMC), Sammon Mapping (SAM), Principal
Component Analysis (PCA), Gaussian Mixtures Models (GMM) and Self Organizing Maps (SOM).
Furthermore, every clustering method is performed alone and the resulting clusters are commented
regardless of the other classifiers results. We also recall that every book has been segmented into 25
several text segments (i.e. 14 for the Quran and 11 for the Hadith) and that there is no prior information
on how could be the general configuration of the clusters (i.e. resulting clustering).
extracted from the two different books: Quran and Hadith respectively, it is quite evident to get
interesting information from the number of obtained clusters and the text segments contained within
each cluster. For instance:
© Copyright 2021-2025 Sayoud.net. All Rights Reserved. 115
a) If we get only 1 cluster, this means that probably the different texts are written by the same
author (i.e. one author);
b) Also, if we get several clusters, but some Quran texts are grouped with some Hadith ones in a
same cluster, this means that some Quran texts were probably written by the Hadith author;
c) However, if two clusters appear in the graphical clustering area and all the Quran texts are
grouped in one cluster and all the Hadith texts are grouped in another distinct cluster, this will
implies that the two books (Quran and Hadith) are written by two different authors (two different
styles).
That is, by exploring the results section and by observing all the clusters and the texts disposition in
those clusters, we easily see that all the obtained results correspond to the third case (i.e. case c). In other
words, all the clustering methods led to two distinct clusters: one cluster containing the Quran texts and
another distinct cluster containing the Hadith texts, in a visual/graphical way.
The visual analytics approach has revealed a lot of information; since it does not only show the
distinction between the author styles, but also shed light on how consistent was that distinction. And
that consistency can be visually estimated thanks to the 3D or 2D separation distance between text
samples. So, the new visualization diagram of the clustering techniques is quite interesting since it
allows seeing how much a particular text sample is far from one author style or near to it.
Finally and statistically speaking, it appears that the two investigated books (Quran and Hadith) possess
two different writing styles and should belong to two different authors, which means that the Quran
could not be an invention of the Prophet.
© Copyright 2021-2025 Sayoud.net. All Rights Reserved. 116
18 Twelfth Series of Experiments: Authorship
Discrimination based on Word Transition Probability
bigrams were found to have some forward and backward probabilities that are specific to only one author
(in a closed set).
That is, in this study, we propose the use of a new set of features based on the Forward Probability
(FWP) and Backward Probability (BWP). This set of features and their normalized form (deeply
described in this paper) is proposed and employed for the first time to the knowledge of the author.
So, it could be interesting to try employing this new set of features in a task of Author discrimination
between the two religious books.
18.1 Probability Computation Procedure
In the following, we describe the required steps to compute the Forward and Backward probabilities
FWP and BWP of the word bigrams.
Let us recall that, in our study, the term Word Bigram represents a couple of successive words as follows:
[Word1 Word2], where Word1 denotes the prefix (1st word) and Word2 denotes the suffix (2nd word).
Now, let us take an example:
Suppose we are interested in the following Arabic bigram ( ), which is written from the right
to the left and which is referred to by [Word1 Word2]. See figure 18.1.
Figure 18.1: Graphical description of the FWP and BWP probabilities
The different steps of the procedure are:
• Compute the probability of Word1 (e.g., ).
• Compute the probability of Word2 (e.g., ).
• Compute the probability of the bigram [Word1 Word2] (e.g., ).
Word1 (
)
FWP=0.75
Word2 (
)
BWP=0.47
© Copyright 2021-2025 Sayoud.net. All Rights Reserved. 117
• Finally, compute the forward and backward probabilities FWP and BWP using the equations
18.1 and 18.2.
In this context, we will denote the bigram probability by p(Wi), the prefix probability by p(Pi) and the
suffix probability by p(Si), so that .
After computation of the different occurrences related to the previous example (see figure 18.1), we get
the Forward Probability (denoted by FWP):
FWP =
(equal to 0.75 in this example)
(18.1)
and, the Backward Probability (denoted by BWP):
BWP =
(equal to 0.47 in this example)
(18.2)
18.2 Probability Normalization Procedure
In the context of transition probability, the sum of the Forward Probability and Backward Probability
is not equal to one. In order to make the sum of the different probabilities equal to one, for a purpose of
normalization, we have proposed the normalized forward and backward probabilities as follows:
NFWP=
(18.3)
and
NBWP=
(18.4)
where NFWP denotes the normalized Forward Probability and NBWP denotes the normalized
Backward Probability.
In this context, NFWP + NBWP = 1 (18.5)
18.3 Selection of the bigrams
18.3.1 Case of limited set of bigrams
The number of the used word bigrams is very large, that is why we limited our selectin to only 20
bigrams. The selected word bigrams, which are investigated in this research work, are given below
(prefix and suffix are put in vertical disposition).
© Copyright 2021-2025 Sayoud.net. All Rights Reserved. 118
Table 18.1 Used bigrams.
18.3.2 Case of unlimited set of bigrams
If one considers all the existing bigrams or at least those that appear at least twice in the text, it will be
more interesting to do the classification by using Machine Learning tools, such as Linear Regression,
Support Vector Machines, Neural Networks, etc. Though it is still possible to employ simple statistical
distances such as Manhattan distance, Cosine distance or Spearman distance, efficiently.
18.4 Experimental Results
18.4.1 Experiments with limited bigrams
The results of the experiments, made with limited bigrams, are displayed in figure 18.2 for the Quran,
and in figure 18.3 for the Hadith. A cumulative histogram representation is used in both figures, where
the black color represents the forward probability and the orange color represents the backward
probability. Note that the light gray color has been used to represent a null probability.
© Copyright 2021-2025 Sayoud.net. All Rights Reserved. 119
Figure 18.2: Graphical representation of the NFWP and NBWP probabilities in the Quran
0% 20% 40% 60% 80% 100%
Quran NFWP & NBWP
Quran Forward Prob Quran Backward Prob
© Copyright 2021-2025 Sayoud.net. All Rights Reserved. 120
Figure 18.3: Graphical representation of the NFWP and NBWP probabilities in the Hadith. For the
case of zero probabilities, we plotted a light gray bar charts. The red dashed line represents the shape
of the intersection line between the NFWP and NBWP bars in the Quran.
According to figure 18.2, there are 4 bigrams from the Quran that have balanced normalized
probabilities (i.e. NFWP NBWP), namely: ,
, ,
While for the Hadith (see figure 18.3), only 1 bigram has a balanced normalized probability, namely:
As for the intersection line, which represents a borderline curve between the NFWP and NBWP bars
(dashed red curve) in figure 18.3, we
st and 2nd bigrams (i.e. , ) and maybe also the 4th and 18th
bigrams. Hence, among 20 analyzed bigrams, only 4 over 20 bigrams have almost similar NFWP and
NBWP (i.e. 20% of the whole bigrams), while 16 among 20 bigrams, namely 80% of the whole bigrams,
present normalized probabilities that are completely different between the 2 books.
18.4.2 Experiments with unlimited bigrams
In these experiments, the 37 text segments extracted from both the holy Quran and Hadith go through
an LOO cross-validation technique, by using the centroid cosine distance for classification.
The obtained accuracy for this classification is 100% for each validation, and the medium cross-
validation accuracy is 100% too, leading to a total separation between the two books and then the two
styles.
0% 20% 40% 60% 80% 100%
Hadith NFWP & NBWP
Hadith Forward Prob Hadith Backward Prob zero prob
© Copyright 2021-2025 Sayoud.net. All Rights Reserved. 121
18.5 Discussion and Conclusion
In this research work, a new approach of Authorship Attribution has been proposed. The new idea
brought by this approach focuses on the probabilistic transitions within a word bigram.
Hence the different statistical parameters, namely: FWP, BWP, NFWP and NBWP, were computed and
employed accordingly as features to perform the identification task and get an attribution decision.
The task of authorship discrimination between the two religious books was ensured, once by
investigating the transition likelihoods of 20 different word bigrams, and another time by investigating
the transition likelihoods of all the word bigrams present in the text.
The excellent results of both experiments show that this type of probabilistic transition feature is
interesting by unveiling very specific values for every author. Furthermore, unlike classical features that
may be topic dependent or common to several authors, the new proposed probabilistic transition features
are not topic dependent.
Furthermore, a graphical assessment based on visual-analytics has been made thanks to the normalized
parameters NFWP and NBWP (figures 18.2 and 18.3), where we visually noticed that the bar charts
displayed for the 2 books are completely different (i.e. visual analytical assessment), and where 80% of
the word bigrams presented different normalized probabilities.
Finally, according to the obtained results (i.e., 100% of correct discrimination), it appears that the two
investigated books should
results stating that they should probably belong to two different Authors. Concerning the new proposed
approach, we can say that it could be quite interesting in author identification provided that the
documents size is large enough.
© Copyright 2021-2025 Sayoud.net. All Rights Reserved. 122
19 Thirteenth Series of Experiments: Authorship
Discrimination based on Deep Learning Technology
In this new research work, we conduct an analysis of authorship discrimination between the Holy Quran
and Hadith. As usually, the primary objective is to check whether the Quran and Hadith could have been
authored by the same Author or not, while the second objective is to leverage the new Artificial
Intelligence based approach (i.e., Deep Learning), to explore this religious enigma.
The global textual corpus, composed of the Holy Quran and a confident part of the Bukhari Hadith, is
divided into segments of 500 words each.
The original aspect of this new work is the use of a deep neural network model based on Long Short-
Term Memory (LSTM), in comparison to two Machine Learning based classifiers: Support Vector
Machine (SVM) and Multilayer Perceptron (MLP).
The different results of authorship classification have shown that the Quran and Hadith were most likely
authored by different Authors, as proved by the high classification accuracy: 100% with the LSTM,
99% with SVM and 99% with MLP.
Interestingly, the results of this study are in total concordance with the previous research works,
reaffirming once again, that the holy Quran could not have been composed or invented by the Prophet,
and which consequently supports the authenticity of the holy Book.
19.1 Dataset
The global corpus is derived from the complete Quran and a certified selection from the Bukhari Hadith.
The Quran, in its entirety, contains over 87,341 tokens, while the selected subset of the Bukhari Hadith
contains more than 23,068 tokens. The global dataset possesses a high degree of consistency, with the
Quran averaging 315 A4 pages and the Hadith part averaging 87 pages.
Since the sizes are not very close between the two books, we have divided the texts into equal-sized
segments. For this research work, we segmented the books into segments of 500 words each (figure
19.1). This segmentation produced a total of 222 text segments, distributed into 175 for the Quran and
47 for the Hadith.
Figure 19.1. Book segmentation into text segments of 500 words each: 175 for the Quran and 47 for
the Hadith.
© Copyright 2021-2025 Sayoud.net. All Rights Reserved. 123
To ensure the consistency of the results, a 5-fold cross-validation technique is employed. Every fold
contains 222/5 44 documents, while every training fold contains 222/5*4 = 177.6 5222×4 178
documents. This technique of cross-validation is used to ensure the integrity of the results.
19.2 Proposed Model based on LSTM
The proposed Deep Learning model is a Sequential model with five layers, as depicted in figure 19.2.
It is composed of the following layers.
• Embedding Layer (Wang, S., 2023), with an input dimension of 5000 and an output dimension
of 64, handling input sequences of limited maximum length. This layer is used for word
embedding where each word is represented as a 64-dimensional vector.
• LSTM Layer (Staudemeyer, R. C. 2023), with 64 units for processing temporal dependencies.
The LSTM (Long Short-Term Memory) is a type of recurrent neural network that is effective
in sequence prediction problems.
• Dense Layer, with 32 neurons using a ReLU activation for non-linear transformation.
• Dropout Layer, which helps to prevent overfitting by randomly setting the outgoing edges of
hidden units to 0 at each update during training time, in order to mitigate overfitting.
• Output Dense Layer, with 2 neurons using Softmax activation for binary classification
probabilities.
Figure 19.2. The proposed Deep Learning model with five layers.
© Copyright 2021-2025 Sayoud.net. All Rights Reserved. 124
19.3 Experimental Results
In this experiment, we used a deep neural network model based on Long Short-Term Memory (LSTM),
without any specific feature extraction and by employing a 5 folds cross validation (figure 19.3).
Figure 19.3. Set of the 5 folders representing the global dataset, with a 5 folds cross validation
rotation.
Remarkably, the LSTM model achieved very high accuracy in all five folds of the cross-validation (see
table 19.1), resulting in an average accuracy of 100%.
Table 19.1 Experiments of Authorship Classification based on LSTM with 5 folds cross validation,
without any feature specification.
Fold
Accuracy
1st Fold
100%
2nd Fold
100%
3rd Fold
100%
4th Fold
100%
5th Fold
100%
Average 5-fold cross-validation
100%
A second experiment using a Support Vector Machine (SVM) in the same dataset, with character
trigrams as features, provided an average accuracy of 99% across a 5-fold cross-validation. Similarly, a
third experiment using a Multilayer Perceptron (MLP) in the same dataset, with character trigrams as
features, also provided an average accuracy of 99% across a 5-fold cross-validation.
These results have shown that all the models are efficient in authorship discrimination between the Holy
Quran and Hadith, while the LSTM high accuracy indicates its superior ability to capture the stylistic
nuances of the books. Once again, these new results support the hypothesis that the Quran and Hadith
should come from different authors.
© Copyright 2021-2025 Sayoud.net. All Rights Reserved. 125
19.4 Conclusion
In a challenging continuation of our previous work, for which the main objective was to check whether
the Quran and Hadith could have been authored by the same Author or not, in this new investigation, we
proposed and applied a deep neural network model based on Long Short-Term Memory in a task of
automatic text classification per author style. This deep architecture was applied on quite small text
segments of 500 words each, comparatively to Support Vector Machines and Multi-Layer Perceptron.
The Deep neural network (i.e., LSTM) has shown two advantages: firstly, it presented better
performances than conventional machine learning classifiers; and secondly it did not require any feature
at its input.
Thus, based on the high classification accuracy achieved by all the models, it can be concluded that the
Quran and Hadith are very likely to have been authored by two different Authors. The LSTM model, in
particular, achieved a very high accuracy of 100%, suggesting a distinct difference in the linguistic
features and stylistic characteristics between the two books, even with small text segments.
This conclusion is well supported by the alignment of these results with our previous research works in
this field. So, once again, the current findings support the authenticity and originality of the holy text,
by refuting the skeptical assertion that the Prophet could have fabricated or dictated the holy Quran.
© Copyright 2021-2025 Sayoud.net. All Rights Reserved. 126
20 Book Analysis based on Embedded Scientific
Knowledge
In the present section, we will cite some new scientific discoveries that were described in the holy Quran
14 centuries ago.
20.1 Scientists talking about the scientific aspect of the Quran
Here are some famous scientists who gave their appreciation on the authenticity of the holy book.
Prof. Jeffrey Lang, Department of Mathematics, Kansas University, USA
scientific notions. Often there appears to be profound similarities. But, more notably, as Bucaille
antiquity that describe or attempt to explain
the workings of nature in that it avoids mistaken concepts.
in the mentality of man, culminating in an age when reason and science would be viewed as the final
criterion of truth. (Ref. J. Lang, Book: Struggling to Surrender, page 37).
Prof. Arthur Alison, Department of Electrical and Electronic Engineering in the University of
London, UK
great. Then I was convinced that Islam is the most proper religion that befits my inborn nature and
conduct. In the heart of my hearts I had felt that there is a God controlling the Universe. He is the
Creator"
that is the revealed religion from the one and only God." (Ref. First Islamic International Conference
on the Medical Inimitability in the Quran, Cairo 1985)
Prof. Zaghloul El-Naggar, Professor of Geology, Head, Committee on Scientific Nations in the
GQPS, Cairo, Egypt.
universe or one of its components, come in a scientific way and with an extreme precision, which proves
that this holy book cannot be a Ref. Zaghloul El-Naggar,
http://www.assabile.com/zaghloul-el-naggar-316/series/wa-yatafakkarun-316).
© Copyright 2021-2025 Sayoud.net. All Rights Reserved. 127
Dr. Gary Miller, Canadian Assist. Professor in Mathematics and former Christian theologian,
Toronto/Canada and KFUPM University Saudi Arabia.
offered by other religious scriptures, in particular, and other religions, in general. It is what scientists
demand. Today there are many people who have ideas and theories about how the universe works. These
people are all over the place, but the scientific community does not even bother to listen to them (Ref.
www.irfi.org/articles/articles_551_600/scientific_approach_to_the_qur.htm).
Prof. Gerald G. Goeringer, Georgetown University, Washington, DC, USA
" No such distinct and complete record of human development, such as classification, terminology, and
description, existed previously...this description antedates by many
traditional scientific literature." (Ref. http://scienceislam.com/scientists_quran.php).
Prof. William W. Hay, University of Colorado, Boulder, Colorado, USA
" I find it very interesting that this sort of information is in the ancient scriptures of the Holy Quran...I
would think it must be [from] the divine being." (Ref. http://scienceislam.com/scientists_quran.php).
Prof. T. V. N. Persaud, University of Manitoba, Winnipeg, Manitoba, Canada
" - You have someone illiterate making
profound pronouncements - amazingly accurate about scientific nature... this is a divine inspiration or
revelation.." (Ref. http://scienceislam.com/ scientists_quran.php).
Prof. E. Marshall Johnson, Thomas Jefferson University, Philadelphia, Pennsylvania, USA
" The Quran describes not only the development of external form, but emphasizes also the internal
stages, the stages inside the embryo, of its creation and development, emphasizing major events
recognized by contemporary science...I see nothing in conflict that ...divine intervention was involved."
(Ref. http://scienceislam.com/scientists_quran.php).
Prof. Alfred Kroner, Institute of Geosciences, Johannes Gutenberg University, Mainz, Germany
" Thinking about many of these questions and thinking where Muhammad came from, he was after all
a Bedouin. I think it is almost impossible that he could have known about things like the common origin
of the universe,...Someone 1400 years ago could not know the heavens and the earth had the same
origin.." (Ref. http://scienceislam.com/scientists_quran.php).
© Copyright 2021-2025 Sayoud.net. All Rights Reserved. 128
Prof. Keith Moore, University of Toronto, Ontario, Canada
"It is clear to me that these statements must have come to Muhammad from God or Allah, because most
of this knowledge was not discovered until many centuries later.." (Ref.
http://scienceislam.com/scientists_quran.php).
Prof. Joe Simpson, Baylor College of Medicine, Houston, Texas, USA
".. It follows, I think, that not only there is no conflict between genetics and religion but, in fact, religion
can guide science by adding revelation to some of the traditional scientific approaches, that there exist
statements in the Quran shown centuries later to be valid, which support knowledge in the Quran having
(Ref. http://scienceislam.com/ scientists_quran.php).
Prof. Yoshihide Kozai, Tokyo University, Hongo, Tokyo, Japan
So, by reading [the] Quran
(Ref. http://scienceislam.com/ scientists_quran.php).
Prof. Tejatat Tejasen, Chiang Mai University, Chiang Mai, Thailand
" During the last three years, I became interested in the Quran.... From my study and what I have learned
from this conference, I believe that everything that has been recorded in the Quran fourteen hundred
years ago must be the truth, that can be proved by the scientific means. " (Ref.
http://scienceislam.com/scientists_quran.php).
Dr Maurice Bucaille, French medical doctor, member of the French Society of Egyptology,
France
The above observation makes the hypothesis advanced by those who see Muhammad as the author of
the
in terms of literary merit, in the whole of Arabic literature? How could he then pronounce truths of a
scientific nature that no other human being could possibly have developed at the time, and all this
without once making the slightest error in his pronouncements in the subject?
The ideas in this study are developed from a purely scientific point of view. They lead to the conclusion
that it is inconceivable for a human being living in the seventh century A.D. to have made statements in
t do not belong to his period and for them to be in keeping
with what was to be known only centuries later. For me there can be no human explanation to the
(Ref. Maurice Bucaille, Book: The Bible, The Qur'an & Science, page 91).
© Copyright 2021-2025 Sayoud.net. All Rights Reserved. 129
20.2 Number of Months and Days in the Quran: "An Enigma"
In the Quran, it is stated that the number of months (per year) is 12, as quoted in the verse [9-36]:
""
Translation: ["The number of months in the sight of Allah is twelve (in a year)- so ordained by Him the
day He created the heavens and the earth"].
But strangely, by counting the occurrence number of the word "Month" in the Quran text, we find
exactly 12 occurrences (cited 12 times in the Quran) see figure 19.1.
Again, if we count the number of "Day" in the Quran text we find exactly 365 occurrences (cited 365
times in the Quran), which is equal to the real number of days per year. Really, it is a fascinating enigma
see figure 19.2.
Figure 19.1 Strange concordance between the occurrences of the word Month and the real number of
months per year.
Figure 19.2 Strange concordance between the occurrences of the word Day and the real number of
days per year.
Really, by observing these two strange coincidences, we do not find any word to say except the fact
that the Divinity origin of the holy Quran is incontestable and evident.
20.3 Earth Rotation in the Quran
It is well-known that the first person who discovered that the earth rotates around its axis
was Copernicus(16th century). Prior to that, scholars thought that the earth was immobile. However,
when we read the holy Quran, we see that it clearly speaks about the earth movement. Hence, in the Ant
verse (verse number 27), Allah says:
-You see the mountains and think that they are immobile and fixed in place while they are
in movement like the clouds. This is the work of God who perfected all things, and verily He is aware of
what you do..
12 occurrences:
Cited 12 times in
the Quran text
There are 12
months per year
Month
365 occurrences:
Cited 365 times in
the Quran text
There are 365 days
per year
Day
© Copyright 2021-2025 Sayoud.net. All Rights Reserved. 130
This verse speaks about the mountains movement although we all imagine them to be still and fixed at
first impression. It also informs us that they are moving like the clouds (see, below, the NASA
image showing the movement of the mountains/earth and a Real photo showing the movement of the
clouds). Consequently, since the scientific community was mistaken before the 16th century, it is clear
that this very ancient information represents a miracle that shows the divine origin of the holy book.
Figure 19.1 Mountains movement seen by satellite (NASA courtesy)
Figure 19.2 Clouds movement
© Copyright 2021-2025 Sayoud.net. All Rights Reserved. 131
20.4 Expansion of the Universe in the Quran
The discovery of Edwin Hubble stating that the Universe is expanding at enormous speed was
revolutionary. He noted that galaxies were all moving away from us, each at a speed proportional
to its distance from us. Today, several NASA spacecrafts continue Hubble's work of measuring the
expansion of the Universe. So, the Universe expansion theory is proved scientifically: it is called
BigBang (see figure 19.3), but the problem is that there are unknown forces at work we do not
understand at all.
Some fascinating evidences on the expansion of the Universe are present in the holy Quran: see the
following verse [51:47]:
-Translation: "And it is We who have built the universe with (Our creative) power; and, verily, it is
We who are steadily expanding it ".
In fact, this verse clearly states that the Universe is continuously expanding, which is in total
confirmation and concordance with that new discovery. That is, how could the Prophet know that
scientific knowledge 14 centuries ago?
Figure 19.3 The Big Bang expansion NASA courtesy.
© Copyright 2021-2025 Sayoud.net. All Rights Reserved. 132
20.5 A Scientific Evidence on the Sun Movement in the Holy Quran
Even though the scientific community thought that the Sun was the center of the Universe and that it
was immobile until the 17th century, it has been clearly stated in the holy Quran, 14 centuries before,
that it does really move in the Universe by respecting a precise orbit that was well defined (by the
Creator); and in perfect concordance with the latest discoveries.
Hence, we can retrieve the related information in the following verse [36:38]:
[36:38](
This verse can be translated into: “And the Sun moves for the fixed/stable course (orbit) assigned for it.
That is the decree/design of the Almighty, the All-Knowing” [36:38].
Figure 19.4: Image of the Milky Way. One can see the orbit of the Sun in yellow. NASA courtesy.
Before the 18th century the Sun has always been considered as the center of the Universe, and then
considered to be at rest (i.e. fixed).
Thus, we can see the heliocentric model of Nicolaus Copernicus in his publication (On the Revolutions
of the Celestial Spheres) in 1543, which was inspired from the assumptions of Aristarchus. His
hypotheses are that the fixed stars and the Sun remain unmoved, and that the Earth revolves about the
Sun in the circumference of a circle, the Sun lying in the middle of the orbit.
Thereafter, Johannes Kepler accepted that theory and published his first two laws about planetary motion
in 1609. In the 17th century, the general idea that the sun was the center of the Universe and immobile
was widely accepted.
Later on, Newton proposed a new heliocentric view of the solar system developed in a modern way
(17~18th century). For Newton, it was not precisely the centre of the Sun that could be considered at
rest, but rather the common centre of gravity of the solar system.
© Copyright 2021-2025 Sayoud.net. All Rights Reserved. 133
Recently (20~21th century), scientists have discovered that the Sun does orbit around the center of the
Milky Way galaxy in a period of about 250 million years, and with a speed of 251 km/s. Moreover, the
Milky Way itself is moving through the Universe within the local group of galaxies (cluster), so that the
estimated global speed of the solar system is about 600km/s (Reference:
http://antwrp.gsfc.nasa.gov/apod/ap050508.html) according to the latest scientific results of NASA.
However, 14 centuries before, in the holy Quran, it has been clearly stated that the Sun does really move
in the Space by respecting a well-defined orbit. So, how could the Prophet give such precise scientific
information in a period of time known by its limited scientific knowledge and instruments, beside the
fact that the Prophet was illiterate?
20.6 About the Embryo description in the Quran
Dr. Keith Moore, Professor of Anatomy and Chairman of the Department, Faculty of Medicine, at the
Highlights of Human Embryology in the Koran and the
Hadith
Dr Moore pointed out that when he studied certain statements in the Quran on this subject,
I was amazed at the scientific accuracy of these statements which were made in the 7th
century CE.
In the Quran (39:6) it is well stated that God created us in the wombs of our mothers in stages. Dr.
Moore saysThe realization that the embryo develops in stages in the uterus was not discussed or
illustrated until the 15th century [https://www.muslimink.com].
Figure 19.5: The embryo stages. GIMS courtesy.
The staging of human embryos was not proposed until the 1940’s, and the stages used nowadays were
not adopted worldwide until a few years ago.”
© Copyright 2021-2025 Sayoud.net. All Rights Reserved. 134
Furthermore, he commentsThe idea that development results from a genetic plan contained in the
chromosomes of the zygote was not discovered until the end of the 19th century. The verse from the
Quran (80:18) clearly implies that the nutfah (i.e. the initial drop of fluid) contains the plan or blueprint
for the future characteristics and features of the developing human being.
The following verse of the Quran (23:12-16) show that there is a gap or lag between two of the early
stages of growth.
Strangely, Dr. Moore saysIt is well-established that there is a lag or delay in the development of the
embryo during the implantation. The agreement between the lag or gap in development mentioned in
the Qur’an and the slow rate of change occurring during the second and third weeks is amazing. These
details of human development were not described until about 40 years ago.
He then concludes by saying that the agreement he has found in the Qurmay help to
close the gap between science and religion which has existed for so many years.
Once again, this scientific discovery shows that the Quran must be from God.
20.7 Description of the Pharaoh's death and the preservation of his
body in the Quran
body preserved after his death (Cairo Museum).
This is the body of Firawn (Rameses II), believed to be the Pharaoh in the time of Prophet Musa (Moses).
His mummy is preserved and is currently on display in the Egyptian Museum, Cairo.
© Copyright 2021-2025 Sayoud.net. All Rights Reserved. 135
The following verses of the holy Quran say (10:90-92):
• We helped the children of Israel cross the sea safely. The Pharaoh and his army pursued the
children of Israel with wickedness and hate until the Pharaoh was drowned. As he was drowning
the Pharaoh said, "I declare that there is no God but the One in whom the children of Israel
believe and I have submitted to the Word of God" (Quran 10:90).
• (God replied), "Now you declare belief in Me! but before this you were a disobedient rebel
(Quran 10:91).
• We will save your body on this day so that you may become evidence (of Our existence) for
the coming generations; many people are unaware of such evidence." (Quran 10:92).
When the Quran was transmitted to the humanity by the Prophet, the bodies of all the Pharaohs, who
are probably related to the Exodus, were in their tombs of the Necropolis of Thebes, on the opposite
side of the Nile from Luxor. At the time however, absolutely nothing was known of this fact, and it
was not until the end of the 19th century that they were discovered (Ref. The Bible, The Qur'an and
Science, by Maurice Bucaille).
As the Quran states in the verse 10:92 (We will save your body on this day so that you may become
evidence - of Our existence- for the coming generations), the body of the Pharaoh of the Exodus was
rescued, and visitors may see him in the Royal Mummies Room-of the Egyptian Museum.
Note that nothing was known at the time of the revelation of the Quran about the mummy of Rameses
II. This fact and the different discoveries reported by Dr Maurice Bucaille confirm that the evidence
given by the holy Quran (on the preservation of the Pharaoh mummy) presents a clear proof on its
truth and its Divine origin.
© Copyright 2021-2025 Sayoud.net. All Rights Reserved. 136
21 About the “Last Prophet” Meaning and Prediction of
a Prophet called Muhammad in the Ancient
Religious Books
According to the holy Quran, it is well known that Muhammad is considered as the last Prophet sent by
Allah (God). In this study we explore and debate that fact fourteen centuries later, by trying to find out
what could be the consequences and conclusions to deduce. Moreover, we try to explore some ancient
prophecies about the apparition of the Prophet Muhammad, which are reported in some ancient religious
books such as the Bible and book of Habakkuk.
prophecy of the Prophet Muhammad from the ancient religious books, which confirm the veracity and
truthfulness of the holy Quran.
21.1 Last Prophet Concept and Truthfulness of the Holy Quran
As we know, in the Quran it is stated that Muhammad (Pbuh) is the messenger of God/Allah and that
he is the last Prophet. In this survey, we will focus on the following verse (33:40):
(33:40)
Translation: "Muhammad is not the father of any man among you, but he is the Messenger of Allah
and the last (end) of the Prophets. And Allah is Ever All-Aware of everything".
In this holy verse, Allah says that Muhammad is his Messenger and not the father of any man among
his companions. He also added that he is the last prophet, which represents a very interesting information
to explore and debate.
In fact, if another new religion, revealed by a new prophet, appeared after the death of Muhammad
(Pbuh), then this verse would present a contradiction.
However, as we know, no new religion has been revealed after the Islam* and no Prophet has been
reported except those who are well known, such as Jesus/Isa, Moses/Musa, Abraham/Ibrahim, etc. (see
figure 1), and then the previous verse does represent a real proof on the truth of the holy Quran.
* The terme of religion refers to a serious revelation from God/Allah with a Divine holy book and which
is effectively followed by a large population of people, such as Islam or Christianity for instance.
© Copyright 2021-2025 Sayoud.net. All Rights Reserved. 137
Figure 21.1: Prophets tree in accordance to Islam. Photo from Islam-beliefs.net (Islam-beliefs, 2023).
21.2 Other significations on the fact that there will be no more
Prophets
The fact that Muhammad is the last Prophet and that there will be no other Prophet after him, involves
the following important points:
• This Prophet is sent for all the humanity, as clarified by the Prophet (Pbuh);
• The holy Quran will be preserved against alterations since there will be no other Divine book
after. This fact has also been cited in the Quran verse (15:9)
,
which can be translated into “Indeed We have sent down the Quran, and indeed We Ourselves surely are its Guardians”.
© Copyright 2021-2025 Sayoud.net. All Rights Reserved. 138
This is one of the miracles of the Quran, since no one has been able to change even one word of
its text, and has remained in its original morphological form since the 6th Century (A. D).
• The end of the world is expected, as reported by the Prophet (Pbuh) »,
which can be translated into “I have been sent and the Hour (end of the world) as these two (fingers)”.
• There is no need to another new revelation or another new religion. This fact can be confirmed
by observing the global religion expansion in the world, as displayed in figure 2, and where one
can notice that Islam is the most growing religion in the world;
Figure 21.2: Estimated percent change in population size by religion. Source: Pew Research center
demographic projection.
According to this study, one can conclude that a lot of information can be deduced from the fact that
Muhammad is the last Prophet and that there will be no other Prophet after him, as discussed in this
paper, but one of the most important points to note here is that this fact represents a real proof on the
truthiness of the holy Quran (since no new religion has been revealed after the Islam).
21.3 Prediction of Muhammad in the Bible and ancient religious
books
• 1st Prediction
The Gospel of Barnabas is attributed to Barnabas and is considered as one of the disciples of Jesus. The
total authenticity of this Gospel has not been established, but it appears quite interesting by explicitly
predicting the Prophet Muhammad (Pbuh) in it. For example, the chapter 163 of the Gospel says:
“… Then Jesus said: ‘So secret is predestination, O brethren, that I say to you, truly, only to one man shall it be
clearly known. He it is whom the nations look for, to whom the secrets of God are so clear that, when he comes into
the world, blessed shall they be that shall listen to his words, because God shall overshadow them with his mercy
even as this palm-tree overshadows us. Yes, even as this tree protects us from the burning heat of the sun, even so
the mercy of God will protect from Satan them that believe in that man.’ The disciples answered, “O Master, who
shall that man be of whom you speak, who shall come into the world?” Jesus answered with joy of heart: ‘He
is Muhammad; Messenger of God, and when he comes into the world, even as the rain makes the earth to bear fruit
-20 -10 0 10 20 30 40 50 60 70 80
Muslims
Christians
Hindus
Jews
Buddhists
Estimated percent change in population size by religion
© Copyright 2021-2025 Sayoud.net. All Rights Reserved. 139
when for a long time it has not rained, even so shall he be occasion of good works among men, through the abundant
mercy which he shall bring ...” (Quran Project, 2023).
According to this chapter, Jesus clearly foretells that there will be another Prophet coming after and his
name is Muhammad, as well expressed in the previous paragraph (sentence in red).
• 2nd Prediction
Pr David Abdu Benjamin Keldani, BD, a Roman Catholic priest of the Uniate-Chaldean sect, wrote the
“… and he shall give you another Parakletos/Periqlytos, that he may stay with you for ever” (John xiv. 16).
Pr Keldani rewrote it in the following explanative form:
“… and he shall send you another apostle whose name shall be Periqlytos, that he may remain with you for ever.”
intended. The two names, one in Greek and the other in Arabic, have precisely the same signification,
(David Abdu Benjamin Keldani, 2007).
• 3rd Prediction
Deuteronomy 33:1-2 (fifth book of the Old Testament) combines references to Moses (Pbuh), Jesus
(probably the village of Sair near Jerusalem) and shining forth from Paran. According to Genesis 21:21,
the wilderness of Paran was the place where Ishmael (Pbuh) settled. In other words, it was in Arabia,
and specifically in Makkah.
(bpuh) to Makkah. The text says:
“He shone forth from Mount Paran; he came from the ten thousands of holy ones, with flaming fire at his right
hand.”
troops. If Muhammad (bpuh), who liberated the city of Paran with 10,000 believing, was not the one
who fulfilled this Biblical prophecy, then who was that prophet? (Al-Rassi, M. S., 2019).
• 4th Prediction
igrated from Paran (Makkah) to be
received enthusiastically in Madinah was none other than Prophet Muhammad (bpuh) (Al-Rassi, M. S.,
2019). See the Habakkuk 3:3 paragraph given below.
“… and the Holy One from Mount Paran. Selah His glory covered the heavens, and His praise filled the earth.[3] His
radiance was like the sunlight; rays flashed from His hand, where His power is hidden.[4]”. [Habakkuk 3:3-4].
So according to the admission of the Bible scholars and the Bible itself in Genesis 21:21 Paran is in
he Muhammed (Pbuh)? (Anthony Matthew Jacob, s. d.).
© Copyright 2021-2025 Sayoud.net. All Rights Reserved. 140
21.4 Conclusion
Throughout this study, we explored the significance of Muhammad being considered as the final prophet
and delved into the philosophical and logical implications of this fact, where we can derive four key
implications and conclusions that strengthen the Quran truthiness.
On the other hand, we tried to explore some ancient prophecies about the Prophet Muhammad apparition
in the ancient holy books, where we can also find four amazing prophecies on the coming of another
prophet from the Arabic peninsula who cannot be anyone other than Muhammad (Pbuh), which
strengthen the Prophet truthiness too.
Consequently, we can conclude that all these facts represent another strong proof on the veracity and
truthiness of the holy Quran and the Prophet Muhammad (Pbuh), leading to a wise and thorough
reflection on the holy words of our Creator.
© Copyright 2021-2025 Sayoud.net. All Rights Reserved. 141
22 Does the Heart have a Control on Mind and
Emotions? A Scientific Evidence Supporting what is
Said in the Holy Quran
During centuries, human beings have been raising a key question on the exact role of the heart: is it
concerned with the mind or is it only a blood pump? And in the latter option, is the brain the main part
of the mind?
According to the holy Quran, the Heart has an important role in mind, thoughts and wisdom (See Alaaraf
and Alhadj chapters). Moreover, during the ancient civilizations it was also admitted that the heart was
strongly linked with feelings and thoughts.
However, in the last centuries, the unique definition of the heart has become only medical and limited
to the function of blood pumping.
Fortunately, recent research works in the field of neuro-cardiology have led to very interesting
discoveries by showing the great role of the heart in mind and feelings. These discoveries, which have
completely changed the definition and role of the heart, represent a real scientific revolution.
22.1 What is the main source of the mind and emotions: the Heart,
the Brain or both?
The famous question, concerning the actual location of the mind center or the feeling center, has not
been solved yet. In the holy Quran, the heart plays a key role in human being behavior, while in ancient
medical science it was supposed to be only the brain. So, is it the heart, the brain or something else?
In this context, Rollin McCRaty, Director of the HeartMath research center, said: “Most of us have been
signals. However, it is not as commonly known that the heart actually sends more signals to the brain
than the brain sends to the heart! Moreover, these heart signals have a significant effect on brain function
influencing emotional processing as well as higher cognitive faculties such as attention, perception,
memory, and problem-solving. In other words, not only does the heart respond to the brain, but the brain
continuously responds to the heart” said the HeartMath Institute Research Director (HartMath, 2022).
Consequently, it appears that the ancient medical definition of the heart, supposed to be a simple blood
pump, is misleading. In fact, several research experiments showed that the heart is also responsible for
many emotional functions.
In fact, for many years scientists studied heart from the physiological side and they considered it only a
blood pumping machine, but starting from the twenty one century and because of the high development
in heart transplantation and artificial heart surgeries, researchers started to notice a strange phenomenon
changes are very deep to the extent that after changing the old heart these changes may affect his
personal believes (Al-Kaheel A., 2022).
Moreover, according to recent research works, in the new field of neurocardiology, scientists have
discovered that the heart possesses its own nervous system (Armour, J. A., 2007) : a nervous network
so sophisticated as to earn the description of an intelligent heart-brain. With mover than 40,000 neurons,
this heart-brain gives the heart the ability to sense, process information, make decisions, and even to
show a type of learning or memory.
© Copyright 2021-2025 Sayoud.net. All Rights Reserved. 142
Again, literature in this research field has shown that the heart communicates with the brain in several
ways: through nerve impulses, via hormones, through pulse waves, and through electromagnetic fields
(McCraty, R., 2015).
cardiac afferent neural signals transmitted to the brain can facilitate or inhibit higher cognitive functions.
In other words, during emotional stress,
discordance communicates some signals to the brain that result in the inhibition of higher brain processes
related to perception, reasoning, and creativity (Fredrickson BL, Branigan C., 2005).
Also, according to Fredrickson, this fact explains why we often cannot think clearly, make careless
mistakes, and have little access to our creativity under stress situation. Hence, these negative emotional
states tend to produce a rigid and limited patterns of thought and action by reducing the possibility to
make wise judgments (Fredrickson BL, Branigan C., 2005).
Biochemically speaking, the heart manufactures and secretes oxytocin hormone, which is involved in
cognition, tolerance, trust, etc. For instance, some previous research works reported that the rat heart is
a site of oxytocin synthesis and release, since this hormone was detected in the four chambers of the rat
heart (Jankowski, M., Hajjar, F., Kawas, S. A., Mukaddam-Daher, S., Hoffman, G., McCann, S. M., &
Gutkowska, J., 1998).
Moreover, the heart produces and secretes several other hormones, such as atrial peptide or atrial
natriuretic peptide, which inhibits the release of stress hormones and influences our motivation and
behaviour (McCraty R, Atkinson M, Tomasino D, Bradley RT., 2009).
One of the important conclusions reported by researchers in this field (HeartMath LLC.
age-old associations of the heart with thought, feeling, and insight
22.2 Heart and Brain citation in the Quran
بلقsingular or plural form, was cited 132 times in the holy
Quran, while the brain was not cited in the holy Quran (to the knowledge of the author) except the term
head in its singular or plural form.
Now, when one reads the holy Quran, we strangely discover that the heart has an important role in
feeling and wisdom, as one can see in the following verses:
a. First verse (chapter 7, verse 179):
179
Translation (chapter 7, verse 179): [Certainly We have winnowed out for hell many of the jinn
and humans: ].
b. Second verse (chapter 22, verse 46):
46
Translation (chapter 22, verse 46: [Have they not travelled through the land so that they may
have hearts by which they may exercise their reason, or ears by which they may hear? Indeed,
it is not the eyes that turn blind, but it is the hearts in the breasts that turn blind!].
© Copyright 2021-2025 Sayoud.net. All Rights Reserved. 143
they have hearts with which they do not
, which involves that the heart plays an important role in faith.
Now, what is meant by Heart in the holy Quran? Is it only an abstract concept without any physical
existence (i.e., representing the mind)? Or is it the real heart organ that is concerned with the
cardiovascular system?
To respond to that question, let us observe the second verse (22:46). By reading the first sentence of this
they may have hearts by which they may exercise their reason
which shows that the heart is a key point of wisdom. However, by reading the second part of the verse,
it is the hearts in the breasts that turn blind
-t. This
Quranic clarification gives an interesting response to the previous question, showing that the mind and
belief are linked to the heart.
Now, by comparing the Quran knowledge on the heart and the new neuro-cardiological discoveries, we
observe a great compatibility. Hence, once again, recent scientific discoveries come together
strengthening the knowledge embedded in the Quran and which was sent down 14 centuries before.
Probably, we will continue to discover more and more amazing concordances between the Quran and
future research discoveries.
22.3 Conclusion
Even though the relationship between the heart, brain and mind is still an enigma, which is not
completely solved or even understood, it appears that the Quran shed light on a lot of questions that
were mistakenly answered by some ancient medical scientists.
The recent research works in neurocardiology have led to very interesting results reinforcing the Quran
theory by showing the important role of the heart in mind and feelings. This discovery, which has
changed the interpretation of the heart role, represents a real scientific revolution.
That is, it appears that the new scientific discovery about the important role of the heart in feelings,
wisdom and reasoning are confirming several Quran verses, which give to the heart a key importance in
belief and decision making.
So, once again and according to the recent important discoveries on this field, it is evident that this
ancient book, dating from the 7th century, could not be a human invention and one cannot find any
explanation on how that knowledge was embedded in the Quran except by considering a Divine
intervention.
© Copyright 2021-2025 Sayoud.net. All Rights Reserved. 144
23 Do Animals communicate with each other? A
Scientific Evidence Supporting what was Revealed
in the Quran
In this survey, we analyze the recent discoveries on animal communication, such as birds, dolphins, ants
and honeybees communication. We also try to see the Quran point of view on the matter by exposing
some pertinent verses reporting a speech/communication related to animals.
The recent works cited in this paper affirm that it does exist a real way of communication between
animals, confirming the information revealed by the holy Quran on the subject.
The cited research works and related results, not only confirm the information embedded in the Quran,
but also lead to an important conclusion about the Divinity of this noble book.
23.1 Animal communication in the Quran
The Quran mentions the language of birds. For instance, in the verse (27:16), Allah (The most Gracious
the most Merciful) says:
Translation: And Solomon (Sulaiman) inherited David. He said, "O people, we have been taught the
language of birds, and we have been given from all things. Indeed, this is evident bounty." (Quran
27:16).
The Quran clearly mentions the language of birds, where it can be noticed that the birds speak with each
other. Some birds like parrots can learn human words. Moreover, some humans, for example villagers
of Kusköy in Turkey are used to communicate with each other by whistles like birds. However, the
strange ability hold by the Prophet Sulaiman to have meaningful conversations with birds (see verse
(27:16)) is amazing (Illias S, s. d.).
Here is a part of discussion between Suleiman (Solomon) and the Hoopoe bird:
*
*
(27:20-22)
Translation: (Solomon) inspected the birds and said, "How is it that I cannot see the hoopoe. Is he absent?
I shall certainly punish him severely or slaughter him unless he has a good reason for his absence." Not
long after, the hoopoe came forward and said, "I have information which you do not have. I have come
from the land of Sheba with a true report. (Quran 27:20-22).
In these verses, we notice a discussion between the Prophet Sulaiman and the hoopoe bird, where the
bird not only understand what was said by Sulaiman, but also responds to him too.
Concerning the Prophet Dawud (David), we find the mention of the birds too:
(34:10)
Translation: And We certainly gave David from Us bounty. [We said], "O mountains, repeat [Our]
praises with him, and the birds [as well]." And We made pliable for him iron (Quran 34:10).
© Copyright 2021-2025 Sayoud.net. All Rights Reserved. 145
More surprisingly Prophet Suleiman (Salomon) was even able to listen to ants communicating with one
another, as cited in the verse (27:17-19):
(27:17-19)
presence in ranks. Until, when they came upon the Valley of Ants, an ant said
smiled, amused at her
speech
me and upon my parents and to do righteousness of which You approve. And admit me by Your mercy
-19)
In the previous verse, Allah (The most Gracious the most Merciful) reports the scenario between
Suleiman and the ant, which was calling her community to enter their habitations to avoid being crushed
by the Prophet and his army. The Prophet was able to understand its speech, which proves that ants have
a developed means of inter-communication.
According to the holy Quran, it is stated that every creature is organized in communities, as it is the case
with human beings. This fact is clearly mentioned in the verse (6:38)
Translation: All the beasts on land and flying birds have different communities, just as you (people) do.
Nothing is left without a mention in the Book. They will all be brought into the presence of their Lord
(Quran 6:38)
23.2 Scientific discoveries about animal communication
Looking for animal behavior is an exciting experience, which shows how little we know about the many
species that exist on earth and how little we know on their mean of communication.
We are still discovering more about the strange and high capabilities that animals possess to
communicate each other.
In the songs and rituals they perform, one can detect a real meaning, which represents a real motivation
in studying animal communication and facts that sustain the most enduring inquiries (Rogers L. J. &
Kaplan G., s. d.).
As mentioned in (Rogers L. J. & Kaplan G., s. d.), most communications occur between members of the
same species (intraspecies communication), but there are cases where one species communicates with
another species (interspecies communication). (Rogers L. J. & Kaplan G., s. d.). We can quote for
instance a communication between a furious dog and a female cat trying to protect its babies.
Scientists from different research fields might agree that the study of animal communication and the
discovery of commonalities of some features of communication between animals, raise the possibility
of considering the human language as only one of the different existing alternative modes of
communication (Rogers L. J. & Kaplan G., s. d.).
© Copyright 2021-2025 Sayoud.net. All Rights Reserved. 146
23.2.1 Dolphins communication
Dolphins employ a highly developed acoustic communication system that uses many pertinent features
associated with the term language (Fulton, J. T., s. d.).
As illustrated by Dr. Denise Herzing, Founder and Research Director of the Wild Dolphin Project
[www.wilddolphinproject.org], dolphin communication employs the same mechanisms and signaling
parameters as human speech except for the frequency range of 3900 and medium used (Fulton, J. T.,
s. d.).
Moreover, Dr Herzing asserts that the dolphin language presents a high entropy (suggesting that it could
even approach some human languages), but without identifying any word in that language (Fulton, J.
T., s. d.).
According to Lammers et al. (Lammers, M. O., Au, W. W., & Herzing, D. L., s. d.), if dolphins pay
attention to the whistles structure with an important associated social role, then the evidence presented
here indicates that there is considerably more to the social acoustic signaling behavior of some species
of dolphins than meets the human ear (Lammers, M. O., Au, W. W., & Herzing, D. L., s. d.).
In figure 23.1 below, one can see a Markov model tree representing the probabilistic sequences of two
dolphin whistles. Numbers in boxes represent whistle types. Percentages and direction of arrows shown
represent the probability of one whistle type following a second whistle type. A curved arrow indicates
the probability that a whistle of one type immediately follows itself. This figure shows a certain
probabilistic structure between whistle types.
Figure 23.1 (McCowan, B., Hanser, S. F., & Doyle, L. R., s. d.): One set of two-whistle sequences
shown as a probability tree based on a Markovian first-order (i.e. Shannon second-order entropy)
analysis. Numbers in boxes represent whistle types. The number of whistles for each whistle type
(WT) included in the diagram were: VVT2=188, VVT7=15, WT162=12, WT5=7, VVT108=5,
WT3=5, WT137=1 (McCowan, B., Hanser, S. F., & Doyle, L. R., s. d.).
According to Ferrer-i-Cancho et al., it appears that the statistical properties of the mapping of dolphins
whistle into meanings is really consistent with the hypothesis that dolphins whistles have some sort of
meaning and that dolphins can communicate through them: the use of a specific whistle type is
constrained by the behavioral context, where it can be shown that these constraints are sometimes shared
by several individuals (Ferrer-i-Cancho, R., & McCowan, B., s. d.).
© Copyright 2021-2025 Sayoud.net. All Rights Reserved. 147
23.2.2 Bird communication
Birds produce a large variety of sounds, from high-frequency whistles to simple hooting. All of these
sounds are used to communicate an information to other members of the same species. However, birds
have another way of communication: birds can also communicate with their bodies through movement
or color for example (Birdfact, s. d.).
Communication between individuals of any species, especially birds, relies on sending and receiving
information in a format that can be understood by both parties. Moreover, the sense of hearing is highly
developed in birds, so it is unsurprising that they use the sound signal to communicate with each other
(Birdfact, s. d.).
Some investigations and analyses that were made on some animal sounds and their studies have provided
evidence for syntax-like structures in their communication systems. In linguistics, syntax represents the
rule of combination of meaningful sounds to form higher-order structures like phrases or sentences
s. d.).
Furthermore, vocal communication in birds can roughly be divided into songs and calls. Although, these
two terms are often used for the same meaning, scientists separate them depending on their real message
and targeted function (Birdfact, s. d.).
An example of human-bird communication can be seen in the following video representing an
interesting human-hawk example of communication: Can I talk to my Hawk?
https://www.youtube.com/watch?v=9mWzwWY1TuM (Mercer D., s. d.).
23.2.3 Ant communication
As other animal species, ants also do communicate between them by forming a strong community
collaboration. We do not exactly know all their means of communication, but it is possible that they
could communicate with different means and in different ways.
Recent research works have shown that ants can communicate via large arrays of pheromones and
possess complex olfactory systems, with antennal lobes in the brain, with up to 500 glomeruli (Hart, T.,
Frank, D. D., Lopes, L. E., Olivos-Cisneros, L., Lacy, K. D., Trible, W., ... & Kronauer, D. J., s. d.).
So, it appears that odors could activate hundreds of glomeruli, which would pose challenges for higher-
order processing. The researchers in (Hart, T., Frank, D. D., Lopes, L. E., Olivos-Cisneros, L., Lacy, K.
D., Trible, W., ... & Kronauer, D. J., s. d.) generated transgenic ants expressing the genetically encoded
calcium indicator in olfactory sensory neurons, and by using two-photon imaging, they mapped
complete glomerular responses to four ant alarm pheromones.
Interestingly, alarm pheromones activated almost 6 glomeruli, and activity maps for the three
pheromones inducing panic alarm in their study species converged on one glomerulus.
Their results showed that ants employ precise and stereotyped representations of alarm pheromones.
Furthermore, they stated that a simple neural architecture is sufficient to translate pheromone perception
into behavioral outputs (Hart, T., Frank, D. D., Lopes, L. E., Olivos-Cisneros, L., Lacy, K. D., Trible,
W., ... & Kronauer, D. J., s. d.).
However, it is also possible that ants communicate by sounds too, and the important role that acoustic
signaling has in ant communication is well established and it is unsurprising that other interacting
species present adaptations that relate to the acoustic characteristics of the host (Schönrogge, K.,
Barbero, F., Casacci, L. P., Settele, J., & Thomas, J. A., s. d.).
For concreteness, one can hear a sample of ants sound in this link:
© Copyright 2021-2025 Sayoud.net. All Rights Reserved. 148
https://exploresound.org/2020/01/ants-have-an-acoustic-world-of-their-own/ (Hickling R., s. d.)
23.2.4 Bee communication
Really, bees do communicate and this fact was proven by scientific evidences. In fact, the Austrian
scientist Karl von Frisch (living from 1886 to 1982) observed that the body movements of foraging bees
on their return to the nest from a food source correlate with its direction and its distance (Von Frisch,
K., s. d.-a) (Von Frisch, K., s. d.-b). Scientists have been amazed by the discovery of this encoded
language, based on the waggle dance, and how it could be effectively used to transmit information about
the localization of remote objects (see figure 23.2) [(Hrncir, M., Barth, F. G., & Tautz, J., s. d.).
Figure 23.2 (BeesWiki, s. d.): Waggle dance is one of the main types of communication methods used
by bees (BeesWiki, s. d.).
complex and we are only beginning to understand a little of the large field of bee communication. In
fact, the bee dance movement is only one chapter of the large story of communication processes that is
used by hundreds of bees belonging to a single colony (Hrncir, M., Barth, F. G., & Tautz, J., s. d.).
23.3 Conclusion
In this survey we saw that the holy Quran mentioned the language of birds and ants. For instance, in the
verse (27:16), the Quran clearly mentions the language of birds, where it can be noticed that the birds
speak with each other.
In the verse (27:20-22), we noticed a discussion between the Prophet Sulaiman and the hoopoe bird,
where the bird understanded what was said by Sulaiman, and even responded to him.
Moreover and more surprisingly, as cited in the verse (27:17-19), Prophet Suleiman (Salomon) was even
able to listen to an ant that was talking to its community.
When we read the Quran, we understand that animals are grouped in well-organized communities. In
fact, according to the holy book, and as it is clearly mentioned in the verse (6:38), it is stated that every
creature is organized in communities, like human beings.
So, the holy Book states that animals, or at least those cited in the Quran, do communicate and speak
with each other in their own language, even if we do not understand their speech.
On the other hand, and through this investigation, we could cite different high-quality scientific research
works related to animal communication, such as dolphins, birds, ants and honeybees, illustrated in the
following publications: (Fulton, J. T., s. d.), (Lammers, M. O., Au, W. W., & Herzing, D. L., s. d.),
(Ferrer-i-Cancho, R., & McCowan, B., s. d.), (Birdfact, s. d.), (Spiess, S., Mylne, H. K., Engesser, S.,
d.), (Mercer D., s. d.), (Hart, T., Frank,
© Copyright 2021-2025 Sayoud.net. All Rights Reserved. 149
D. D., Lopes, L. E., Olivos-Cisneros, L., Lacy, K. D., Trible, W., ... & Kronauer, D. J., s. d.) , (Hickling
R., s. d.), (Von Frisch, K., 1965-b) (Von Frisch, K., 1965-a) and (Hrncir, M., Barth, F. G., & Tautz, J.,
2005), and which explicitly show that those animals do communicate within their community and do
possess a real organized way of communication with a specific language, as seen in sections 23.2.1,
23.2.2, 23.2.3 and 23.2.4.
Consequently, the scientific discoveries in this research field clearly confirm the main concept of animal
communication revealed in the holy Quran, and which was reported 14 centuries before.
Now, the rising question would be: Is it possible that the holy Quran, with all that embedded knowledge,
could be written by an illiterate human being from the 6th century?
The response is obviously: No. Moreover, it is not surprising to see many other scientific discoveries
sustaining the Divine origin of this holy book.
© Copyright 2021-2025 Sayoud.net. All Rights Reserved. 150
24 Effect of the holy Quran in Soul Appeasement and
Treatment of Anxiety: An experimental Evidence on
the Divinity of the Book
The holy Quran is a fascinating book that creates several positive effects on the reader. Such impacts
have been reported by a lot of researchers and can even be felt by reading the holy book either in its
original version or in its translated version.
In this paper, we try to expose some related works on the appraisement of the Soul and on the reduction
of anxiety, based on the reading or listening of the holy Quran.
The reported results of those research works have unanimously shown that the holy Quran does have a
real impact on the reduction of anxiety by treating different negative psychological feelings and
depressive disorders. Moreover, it provides hope, confidence and motivation.
24.1 Experimental studies
Four experimental studies were reported by several researchers to try evaluate the effect of holy Quran
recitation on the reduction of anxiety. Those studies are described as follows:
24.1.1 Study 1
The first study was conducted in 2016, with the objective of determining the effect of the Quran
recitation on mental health of the medical staff of Mazandaran University of Medical Sciences.
participants were randomly distributed into two groups (40 participants in control and experimental
group). Experimental group listened to some verses of the Holy Quran for 3 months at the beginning of
Results showed that the mean of mental health and all its domains, after hearing the verses of the Quran,
in experimental group (listening to the Quran) was higher than the control (not listening to the Quran)
group (p <0.05) (Darabinia M., Gorji A.M.H., Afzali M.A., 2017).
© Copyright 2021-2025 Sayoud.net. All Rights Reserved. 151
Figure 24.1: The mean of mental health before and after the playing verses of the Quran (Darabinia,
2017). We can notice that the mental health is better after Quran recitation (see difference between
histograms green and blue) (Darabinia M., Gorji A.M.H., Afzali M.A., 2017).
Their study revealed the positive effect of hearing the Quran on the mental health of participants. As a
conclusion and on the basis of their findings; it can be deduced that hearing the Quran recitations
improves the mental state of people.
Furthermore, regarding the close connection of the teaching staff with the students of the University
where the experiments were done, they showed that hearing the Quran can make the staff to feel
more satisfied and do their tasks with more optimism; thus, the students will be pleased as well. As
a consequence, they recommended the use of Quran recitations to reinforce positive emotions and
psychological comfort for the University staffs (Darabinia M., Gorji A.M.H., Afzali M.A., 2017).
24.1.2 Study 2
Another review study was performed by Ashraf Ghiasi on articles published between January 1990 and
September 2017. Several online databases including Scopus and Google Scholar were searched with the
k of bias across all included studies was assessed
In their study, the authors reported that from 973 articles found in the initial search, 28 randomized
controlled trials and quasi-experiments were selected for the systematic review. Also, In most studies,
State-Trait Anxiety Inventory was used to measure
The results of this review revealed a positive effect of listening to Holy Quran recitation in
reducing anxiety in various settings (Ghiasi A., Keramat A., 2018).
The current evidence indicates that listening to Holy Quran recitation is a useful non-pharmacological
treatment for reducing anxiety. However, due to the limited number of studies in this area, further
research is needed to obtain more accurate evidence (Ghiasi A., Keramat A., 2018).
© Copyright 2021-2025 Sayoud.net. All Rights Reserved. 152
24.1.3 Study 3
Another different research work evaluated the effect of Quranic therapy on psychological diseases. The
experiments have been conducted on 121 patients from both genders. There were different sessions with
the patients, who were given some verses from the Holy Quran for listening during a specific time.
Thereafter, every patient was given a remedy program. This study aimed to measure the effectiveness
of patients to receive treatment through Quran. The results of the effectiveness factor came after ability
and willingness and gave a result of 92.6% for those who support the fact that the Quran has a significant
healing influence. The authors report that some of the patients who regularly attended Quranic
therapy sessions have been successfully cured; and 81.8% of the sample believe that Quranic therapy
supports their health needs. They also concluded by stating that this study has empirically proved that
the sound of the Holy Quran is an effective treatment for spiritual and psychological issues (Saged,
A.A.G., Mohd Yusoff, M.Y.Z., Abdul Latif, F. et al., 2020).
24.1.4 Study 4
In 2022, Gavgani et al. (Gavgani Z., V., Ghojazadeh, M., Sadeghi-Ghyassi, F., & Khodapanah, T., 2022)
tried to evaluate the effects of listening to Quran recitation on reducing preoperative anxiety, since such
anxiety is a very common unpleasant reaction among patients waiting to undergo a surgery. A systematic
review, for collecting the data, was performed in Medline, EMBASE, Cochrane Library, PsycINFO, Arab
World Research Source, and other relevant databases.
Randomized trials about the effects of listening to Quran on preoperative anxiety reduction in elective surgery
were selected without language or date restriction.
2 index with 50% threshold were used for calculating the heterogeneity
and inconsistency index. Furthermore, subgroup analysis was conducted based on the surgery type and the
funnel plot was used to assess the possibility of publication bias. Basically, twelve studies were included in
the qualitative synthesis and nine were included in the quantitative synthesis.
The meta-analysis showed a significant anxiety reduction with listening to Quran recitation. The
2
So, the analysis showed that listening to Quran recitation reduces anxiety in both major and minor surgeries.
The findings of this statistical investigation indicated that listening to Quran recitation can be
considered as a non-invasive and peaceful intervention to reduce preoperative anxiety among patients
waiting to undergo a surgery (Gavgani Z., V., Ghojazadeh, M., Sadeghi-Ghyassi, F., & Khodapanah, T.,
2022).
24.1.5 Summary of the four studies
In study 1, the researchers showed that hearing the Quran can make the staff to feel more satisfied and
do their tasks with more optimism.
In study 2, the results revealed a positive effect of listening to Holy Quran recitation in reducing anxiety
in various settings.
In study 3, the authors reported that some of the patients who regularly attended Quranic therapy
sessions have been successfully cured.
In study 4, the results indicated that listening to Quran recitation can be considered as a non-invasive
and peaceful intervention to reduce preoperative anxiety.
Hence, the four experiments have shown that listening to the recitation of the Quran does have a real
positive impact on the soul appraisement and anxiety reduction.
© Copyright 2021-2025 Sayoud.net. All Rights Reserved. 153
24.2 Quran effect in coping with anxiety
The way people cope with anxious events represents a major factor in whether or not their health is
affected. For Instance, if we take the example of divorce, most people would say that divorce is one of
the most destructing life crises, since it affects all the family and generates a serious stressful
environment (Ayad, A., 2008).
That is, let us look at the Talaq chapter (verses 2 and 3) in the Quran. Reading these verses makes you
feel as if a merciful hand is caressing you and giving you hope for the future. In fact, the verses are filled
with positive attitude - assurances from Allah that things will get better, that there is a foreseeable end
to the current problems and sadness (Ayad, A., 2008).
(65:2-3)
Translation: So when they are about to reach their appointed term, hold them back with kindness or
separate them with kindness, and make two just men among you as witnesses, and establish the testimony
for Allah; with this is advised whoever believes in Allah and the Last Day; and whoever fears Allah –
Allah will create for him a way of deliverance (2). And will provide him sustenance from a place he
had never expected; and whoever relies on Allah – then Allah is Sufficient for him; indeed Allah will
accomplish His command; indeed Allah has set a proper measure for all things (3). [65:2-3]
Amazingly, by only reading these verses, with a sincere belief in Allah's promise, power, mercy and
wisdom, is sufficient to reduce stress and give hope, which will help the depressed persons to cope with
an anxious situation, even in the most difficult situations (Ayad, A., 2008).
In a psychological point of view, the cognitive appraisal as introduced by Professor Richard Lazarus
(Lazarus, R. S., & Folkman, S., 1984), could describe how different changes and various encountered
circumstances influence individuals. So, some people perceive any problem as menacing and stressful;
while others approach their problem with a fighting spirit, favoring adjustment and adaptation. Thus,
perceiving stressful situations as harmful complicates our ability to cope with these situations. On the
other hand, seeing them as challenging enables us to deal efficiently with the events (Ayad, A., 2008).
However, for believers, the way of coping with emotions depends primarily on the degree of faith. A
deep trust in Allah, associated with the fact that this world is transient, can give a real strength and
feeling of peace and satisfaction. A true believer never falls into despair, since Allah is present and since
He promises to reward patience in this life and in the hereafter (Ayad, A., 2008). See the following verse
of the Quran:
9:51
Translation: Say: "Nothing will happen to us except what Allah has decreed for us: He is our
protector": and on Allah let the Believers put their trust. (Quran 9: 51)
In this way, the fact of reading the Quran, associated with a great faith in Allah, can influence the
response to a stressful situation, in changing the fear or anxiety into calm, peace and hope. It also highly
contributes in improving the coping attitude by producing a relaxing feeling and by calming down the
stress.
© Copyright 2021-2025 Sayoud.net. All Rights Reserved. 154
24.3 Discussion
In this scientific survey, we noticed some real positive effects of the holy Quran (reading, listening or
recitation) on the appeasement of anxiety and the improvement of motivation in stressful or difficult
situations.
Several experimental investigations were made on real patients, as reported by Darabinia in the 1st study
(Darabinia M., Gorji A.M.H., Afzali M.A., 2017), by Ghiasi in the 2nd study (Ghiasi A., Keramat A.,
2018), by Saged in the 3rd study (Saged, A.A.G., Mohd Yusoff, M.Y.Z., Abdul Latif, F. et al., 2020) and
by Gavgani in the fourth study (Gavgani Z., V., Ghojazadeh, M., Sadeghi-Ghyassi, F., & Khodapanah,
T., 2022). Strangely, the four experiments have shown that listening to the recitation of the Quran does
have a real positive impact on the soul appraisement and anxiety reduction, and these discoveries were
scientifically proved by statistics.
That is, what could be the secret in such spiritual strength? Is it only the effect of words? But what types
of words could influence so deeply the soul and treat many psychological pains? Sincerely, we do not
see any possible cause except the fact that the holy Quran should be sent down by a Super Power Creator
(Allah, praise be upon Him) who embedded all the necessary treatment required in such situations in
His holy book.
© Copyright 2021-2025 Sayoud.net. All Rights Reserved. 155
25 Statistical Investigation on Ancient Quran Folios:
Case of Birmingham and Sanaa Parchments
In 2014, some scientists of the University of Birmingham discovered that four folios containing some
ancient Quran manuscripts can be dated sometime between 568 and 645CE. This means that the animal
from which the skin was taken was living sometime between these dates. Similarly, in 1965 heavy rains
damaged the roof construction of the Western Library in the Great Mosque of anaa, where scientists
discovered some Quran folios that should belong to the period between 578 CE and 669 CE. This means
that those Quran manuscripts are probably
In this investigation, we want to check whether those ancient texts are similar to the present Quran or
not, and if the two ancient manuscripts, discovered in Birmingham and Sanaa, contain similar text or
not.
The first results, based on character analysis and word analysis, have shown that the two old folios are
very similar to their corresponding part contained in the present Quran (Uthmanic compilation). So, it
appears that the morphological skeleton of the analysed Quran text has been safely preserved during the
last 14 centuries.
25.1 Introduction
A brief description and history on the two old parchments, namely: Birmingham Quran and Sanaa
Quran, are depicted below.
25.1.1 Introduction on the Birmingham Quran manuscript
The Birmingham Quran manuscript consists in four pages made of parchment, written in ink, and
containing parts of chapters 18, 19 and 20 of the holy Quran (Fedeli A., 2015). The manuscript forms
Cadbury Research Library (Birmingham University, 2015).
(Hopwood D., 1961).
The collection came to the University of Birmingham in the late 1990s. Concerning the palaeographic
aspect of the manuscript (Titled Hejazi text), the handwriting geometry suggests that it may have been
created in the Hejaz area in the west of the Arabian Peninsula, which includes the sacred cities of Mecca
and Medina. In fact, there are several old manuscripts dating from the first centuries after the Hijra,
where we can clearly see the difference in the palaeographic style (Awwad K., 1982). The palaeography
can give a quite good estimation on the probable date of the manuscript, but the radiocarbon dating is
usually more accurate with regards to the parchment dating. This last technique is widely used in
archaeological dating (Taylor R.E. and Aitken M.J., 1997).
Thus, the radiocarbon analysis, made at the Radiocarbon Accelerator Unit of Oxford University
(Ramsey B. C., Higham T. F. G., Brock F., Baker D., & Ditchfield P., 2009), yielded the following
technical dating results (Higham T. F., Bronk Ramsey G. C., Chivall D., Graystone J., Baker D.,
Henderson E., Ditchfield P, 2018) (Birmingham University, 2016):
▪ Radiocarbon Reference: OxA-29418
▪ Radiocarbon Result: 13 C = -21.0 1456+/-21
▪ Calibrated Date Range: 95.4% probability to be between 568 and 645 AD
Hence, the manuscript has been radiocarbon-dated by the University of Oxford (Radiocarbon
Accelerator Unit) to the date range of 568645 CE with a 95.4% degree of confidence. The radiocarbon
result means that the animal from which the skin was taken was living sometime between these specific
dates. This places the discovered parchment close to the death of the Prophet who lived between 570
and 632 CE.
© Copyright 2021-2025 Sayoud.net. All Rights Reserved. 156
Some researchers argued that the manuscript is among the earliest written textual document of the Quran
known to survive, which was written few years after the Prophet death. They also claim that it should
probably be the oldest Quran manuscript in the UK.
In this investigation, we try to check whether the ancient text is similar to the corresponding Sanna
Quran part and to the present Uthmanic Quran by means of comparative analysis.
25.1.2 Introduction on the Sanaa Quran manuscript
In 1965 heavy rains damaged the roof construction of the Western Library in the Great Mosque of anaa.
A single window was discovered to contain a substantial cache of used Arabic manuscripts, almost all
being ancient manuscripts of the Quran spanning the first few Islamic centuries. A Colloquium on the
Islamic City organised by the World of Islam Festival Trust, sponsored by UNESCO, was held at the
University of Cambridge, in July 1976. Drawing a wide variety of experts from both the Muslim and
non-Muslim world, a number of specific research activities were recommended, amongst which was
highlighted the pressing need to conserve the rich corpus of Quranic texts discovered in the Mosque of
anaa (UNESCO, 1980).
The anaa collection was made known to the general public with the publication of Maif
1985, an exhibition catalogue presenting some of the findings of the project. A single palimpsest folio
-Makhe., DAM 01-27.1), folio 21a according to
script and its contents. The folio was tentatively dated to the first half of the 1st century of hijra (Sanaa,
1985).
In 2010 Sadeghi and Bergmann had published their article analysing the four auction folios, specifically
the Sotheby’s 1993 / Stanford 2007 folio, where details were given of a radiocarbon study corroborating
the early date already assigned to the manuscript. Analysis was done at the Accelerator Mass
Spectrometry Laboratory of the University of Arizona (Sadeghi, B., & Bergmann, U., 2010). According
belonging to the period between 578 CE and 669 CE.
In our investigation, we are particularly interested in the parchment referenced by the reference
029006B, by making a statistical comparison of this manuscript with its corresponding part in
Birmingham folios and the present Uthmanic compilation.
25.1.3 Goal of this work
We try to conduct a statistical analysis for evaluating the difference, if any, between the discovered
parchments and the present compilation of the Quran. The used discrimination technique is based on the
computation of characters and words that are different, within the manuscripts.
25.2 Notes on the ancient Arabic handwriting
Over the time, a great variety of manuscript copies of the Quran survived, along with commentaries
written by several ancient scholars. Furthermore, a large exploration of the ancient Arabic manuscripts
shows how the Quran, in terms of calligraphy, evolved over time in different regions of the world
(Ansorge, C., 2011).
25.2.1 Hijazi script
The earliest Quran manuscripts were probably written in the 7th century. Its script represents a very
early Quran text written on a single large sheet of parchment and folded in half. This primitive script is
m the Arabian Peninsula around Mecca and Medina. The
few dots and other markings indicating pronunciation or pauses (Ansorge, C., 2011).
© Copyright 2021-2025 Sayoud.net. All Rights Reserved. 157
25.2.2 Diacritics and rasm dots
The new Arabic writing structure has undergone a gradual improvement, which may be noticeable in
modern printed Quran books (Denffer A.V., 2009). The first change, in the writing structure, was the
addition of the diacritical marks (tashkeel), corresponding to vowels. The second one was the addition
of dots, for discriminating between some close consonants.
So, the ancient Arabic characters are a bit different from what we are used to write nowadays. In fact,
as displayed in figure 25.1, there is a correspondence between the ancient characters (at the bottom) and
the new equivalent ones (at the top).
Figure 25.1: Ancient Arabic characters and their corresponding recent Arabic characters.
One of the particularities of the ancient Arabic scripture is the fact that the text was not vocalised and
did not contain Rasm dot marks. For example, in the ancient script, the word Alaiha (
) was written
as follows: .
25.2.3 Note on the "silent alif"
The natural elongation, called: Al-Madd Al-tabeee, is the act of elongating the sound of the three basic
vowels (Muqith M. A., 2011), consisting in stretching the sound of a character to two basic ones or more.
However, the Ancient Arabic script was not vocalised and often not accompanied by elongation marks.
For the case of the vowel A (fatha in Arabic), the corresponding elongation mark is called “silent alif”.
But we have noticed that in the ancient manuscripts, the silent alif (elongated A) was not very used as it
is the case in the recent Arabic text. In fact, prior to the orthographic reforms carried out under the
auspices of the Umayyad caliph Abd al-Malik (AH 80), the Arabic script did not contain a lot of woyels,"
i.e., there were no short vowels and no grapheme to represent the hamza or the long alif, example: the
verb qaala (
) in the modern Arabic script, was often written qala (
) without silent alif in the
ancient Arabic one.
That is why, it is not surprising to see the ancient Quran manuscript without diacritics or without
elongation marks, since most of these marks were invented several centuries afterward, probably under
the auspices of the Umayyad caliph Abd al-Malik in 80 AH (Powers D. S., 2011).
25.3 Analysis of the Birmingham Quran
In this investigation we will make a comparative analysis on the four ancient Birmingham folios with
regards to their corresponding verses in the present Quran (Hafs Recitation). Our discrimination
technique is based on the computation of characters and words that are different, within the manuscripts.
The two first folios correspond mainly to chapter 18 (Al-Kahf), and the two last folios correspond mainly
to chapter 20 (Taha) and the end of chapter 19 (Maryam).
The following table (table 25.1.a) consists of two columns; the left one contains a part of Birmingham
Quran folio and the right one contains the corresponding verses in the present Quran (Hafs recitation
with Uthmanic rasm) without vocalisation (i.e. without diacritics).
© Copyright 2021-2025 Sayoud.net. All Rights Reserved. 158
Table 25.1.a: Example of comparison between a part of Birmingham Quran and the present Quran
without vocalisation
Table 25.1.b: Example of comparison between a part of Birmingham Quran and the present Uthmanic
Quran
By comparing all the folios with the present Uthmanic rasm, we have obtained the following statistics
that are presented in table 25.2.
Table 25.2: Statistics of all Birmingham folios
Manuscript
Number
of verses
Number
of lines
Similarity
in terms of
words
Similarity
in terms of characters
without considering
alif”
Similarity
in terms of characters
by considering the
alif”
Birmingham Folio 1
7
24
100%
100%
98.08%
Birmingham Folio 2
9
23
100%
100%
98.74%
Birmingham Folio 3
19
23
100%
100%
99.5%
Birmingham Folio 4
28
23
100%
100%
98.6%
All Folios
63
93
100%
100%
98.71%
Note: The similarity in terms of words (or characters) is equal to: 100% - the number of different words (or characters) that
are different in percent.
As one can see in table 25.2, which represents a statistical comparison between this ancient Quran folios
and the present one compiled by Uthman, it appears two important conclusions:
▪ The two analysed text documents are similar in terms of words (similarity of 100%);
▪ The two analysed text documents are quite similar in terms of characters (similarity of about
100% without considering the sildent alif and about 99% by considering the sildent alif);
Consequently, and since the ancient Birmingham scripture was found to be morphologically similar to
the present holy scripture, it appears that the skeletal morphology of the Quran, for the investigated
chapters, has been safely preserved during the last 14 centuries.
Present Quran (old rasm
without vocalisation)
Birmingham Quran
Present Uthmanic Quran
Birmingham Quran
© Copyright 2021-2025 Sayoud.net. All Rights Reserved. 159
25.4 Analysis of the Sanaa Quran
In this investigation we will make a comparative analysis between the present Uhmanic Quran and its
available corresponding part found in Sanaa. We should note that due to the limited available data in
Sanaa Quran, we could only make a comparison from the end of chapter 19 and the beginning of chapter
20 in in the Sanaa Quran. Our discrimination technique is based on the computation of characters and
words that are different, within the manuscripts. Once again, the diacritics and modern writing rasm
have been removed to put the two texts in the same writing conditions (i.e. diacritics removal).
25.4.1 Statistical analysis of the Sanaa folio
In summary, by taking the Sanaa Quran and comparing this last one with the present Uthmanic Quran
without vocalisation, we have obtained the following statistical results.
Table 25.3: Similarity between the ancient Sanaa folios and the present Uthmanic Quran
Comparison between the ancient Sanaa folios
and the present Uthmanic Quran
Similarity in %
Similarity in terms of words
100 %
alif”
100 %
alif”
99.07%
Note: The similarity in terms of words (or characters) is equal to: 100% - the number of different words (or characters)
that are different in percent.
So, as one can see in table 25.3, all the similarity ratios are almost equal to 100%, showing that the old
Sanaa parchment is almost identical to the present Quran in terms of skeletal morphology.
25.4.2 Comparison between Sanaa and Birmingham folios
In this investigation, we make a comparison between the two ancient folios. However, due to the
limitation in the available data, the comparison is made between the verses [19:91] and [20:39].
The statistical results of this comparison are represented in table 25.4.
Table 25.4: Similarity between the Sanaa Quran folios and Birmingham Quran folios
Comparison between the ancient folios and the current holy Quran
Similarity in %
Similarity in terms of words
100 %
alif”
100 %
alif”
99.84%
Note: The similarity in terms of words (or characters) is equal to: 100% - the number of different words (or characters)
that are different in percent.
We represented the images of these ancient folios in figure 25.2, the left side contains the 3rd and 4th
Birmingham Quran folios, while the right side contains its corresponding verses in Sanaa Quran.
© Copyright 2021-2025 Sayoud.net. All Rights Reserved. 160
Figure 25.2: Comparison between Birmingham Quran (left) and Sanaa Quran (right), corresponding to
the end of chapter 19 and beginning of chapter 20.
By comparing the two old parchments, one can notice that the different similarity measures are almost
equal to 100%, showing that the two parchments, Sanaa and Birmingham manuscripts, present a very
high similarity, despite their different geographical origins.
25.5 Discussion
In summary, during this investigation, we have undertaken a comparative analysis between the
Birmingham Quran folios, Sanaa Quran folios and their corresponding verses in the Uthmanic
compilation of the present Quran, which correspond to the end of chapter 19 and the beginning of chapter
20, where the comparative analysis is based on characters and words. This investigation is important
and interesting because the parchments of Birmingham and Sanaa Quran, have been carbon-dated to the
first century of Hijra.
Based on the obtained statistical results, we have identified two significant conclusions:
▪ The two analysed manuscripts: Birmingham and Sanaa folios, appear similar in terms of
characters and in terms of words;
▪ The ancient texts (end of chapter 19 and beginning of chapter 20) appear identical to the
corresponding present Uthmanic Quran, with regards to the skeletal morphology of the analysed
texts.
© Copyright 2021-2025 Sayoud.net. All Rights Reserved. 161
Consequently, the old parchments corresponding to chapter 19 and 20, show that the consonantal
morphology of the Quran has been safely preserved for those chapters, during the last 14 centuries
without alteration.
That is, if the radiocarbon dating is quite accurate, we can say that this new discovery confirms that the
present holy book should represent an authentic copy of the first original Quran that was recited by the
Prophet fourteen centuries ago.
From another vision, and by reading the verse (15:9) of the Quran: «
», it is
stated that the holy Scripture is protected and preserved by His Creator; which strengthens the result of
this scientific investigation, at least for the investigated chapters.
© Copyright 2021-2025 Sayoud.net. All Rights Reserved. 162
26 General Conclusion and Discussion
In this survey, we tried to see whether the Quran could be a simple invention of the Prophet (i.e. written
by the Prophet) or really a book from God (i.e. a divine book sent down by Allah) as claimed in the
Islamic religion.
This survey has to be considered as a pure scientific investigation without any form of theological or
ideological point of view. Also, the author does not discuss his personal beliefs on the subject, but only
what the scientific results of this investigation show.
Again, it is important to recall that the consequences of this research work could be very heavy, that is
why we should comment the different results with care, rigor and objectivity.
So, in this second edition of the book, we can find a brief description of the two analysed books
(i.e.Quran and Hadith), we can also find the description of the 13 series of experiments that were
conducted during this investigation and some new scientific knowledge that were embedded in the
Quran.
The 13 series of experiments are depicted as follows:
• 1st Series of Experiments, in chapter 7:
Global Analysis
This is a global authorship discrimination
• 2nd series of experiments, in chapter 8:
Big Segments based Segmental analysis
This is a segmental authorship discrimination
• 3rdseries of experiments, in chapter 9:
Automatic authorship attribution with several
features and several classifiers
• 4thseries of experiments, in chapter 10:
Short Segments based Segmental Authorship
Attribution
• 5th Series of Experiment, in chapter 11:
Stylometric Comparison between the Quran and
Hadith based on Successive Function Words
• 6th Series of Experiment, in chapter 12:
Authorship Identification of 7 Books A Fusion
Approach
• 7th Series of Experiments, in chapter 13:
Authorship Discrimination using the Leave-One-Out
Validation
• 8th Series of Experiments, in chapter 14:
Authorship Discrimination based on Gaussianity and
Interpolability
• 9th Series of Experiments, in chapter 15:
A Mysterious Numerical Structure in the Quran
making it different from other Human Books
• 10th Series of Experiments, in chapter 16:
Authorship Attribution based on the Interrogative
Form
• 11th Series of Experiments, in chapter 17:
Investigation on the Quran/Hadith Authorship Using
Visual Analytics Approaches
• 12th Series of Experiments, in chapter 18:
Authorship Discrimination based on Word
Transition Probability
• 13th Series of Experiments, in chapter 19:
Authorship Discrimination based on Deep Learning
Technology
© Copyright 2021-2025 Sayoud.net. All Rights Reserved. 163
As one could notice in all the results reported in those 13 series of experiments, the conclusions are
style of the Hadith (i.e. Prophet), which leads to the following major conclusion: The Quran could not
be written or invented by the Prophet Muhammad. Hence, the claim that the Quran was sent down by
God (Allah) is widely strengthened and appears to be true.
At the end of this book, more specific research works with regards to some important information and
knowledge embedded in the holy Quran, have been reported. The specific works discussed are presented
and commented in the following chapters:
• Chapter 20:
Book Analysis based on Embedded Scientific
Knowledge
• Chapter 21:
a Prophet called Muhammad in the Ancient
Religious Books
• Chapter 22:
Does the Heart have a Control on Mind and
Emotions? A Scientific Evidence Supporting what
is Said in the Holy Quran
• Chapter 23:
Do Animals communicate with each other? A
Scientific Evidence Supporting what was Revealed
in the Quran
• Chapter 24:
Effect of the holy Quran in Soul Appeasement and
Treatment of Anxiety: An experimental Evidence on
the Divinity of the Book
• Chapter 25:
Statistical Investigation on Ancient Quran Folios:
Case of Birmingham and Sanaa Parchments
•
several new scientific knowledge that were embedded in the Quran and commented by several
famous researchers, and which have further confirmed this conclusion by showing that the holy
book could not be written by a human being, but it should belong to a super-power intelligence
that has an extreme power and very large scientific knowledge (i.e. Allah/ God).
• In chapter 21, titled:
Muhammad in the Ancient Religious Books, we explored the significance of Muhammad being
considered as the final prophet and delved into the philosophical and logical implications of this
fact, where we could derive four key implications and conclusions that strengthen the Quran
truthiness. On the other hand, we tried to explore some ancient prophecies about the Prophet
Muhammad apparition in the ancient holy books, where we could also find four amazing
prophecies on the coming of another prophet from the Arabic peninsula who cannot be anyone
other than Muhammad (Pbuh), which strengthen the Prophet truthiness too.
• In chapter 22, titled: Does the Heart have a Control on Mind and Emotions? A Scientific
Evidence Supporting what is Said in the Holy Quran, we cited some recent research works in
neurocardiology that have led to very interesting results reinforcing the Quran theory by
showing the important role of the heart in mind and feelings. This discovery, which has changed
the interpretation of the heart role, represents a real scientific revolution. That is, it appears that
the new scientific discovery about the important role of the heart in feelings, wisdom and
reasoning are confirming several Quran verses, which give to the heart a key importance in
© Copyright 2021-2025 Sayoud.net. All Rights Reserved. 164
belief and decision making. So, once again, it is evident that this ancient book, dating from the
7th century, could not be a human invention.
• Do Animals communicate with each other? A Scientific Evidence
Supporting what was Revealed in the Quran, we could see that the holy Quran mentioned the
language of birds and ants. In the verse (27:20-22) for instance, we noticed a discussion between
the Prophet Sulaiman and the hoopoe bird, where the bird understood what was said by
Sulaiman, and even responded to him. So, according to the holy book, it is clearly stated that
every creature is organized in communities, like human beings. So, the holy Book states that
animals, or at least those cited in the Quran, do communicate and speak with each other in their
own language, even if we do not understand it. Furthermore, and through the cited
investigations, we could see different scientific research works related to animal
communication, such as dolphins, birds, ants and honeybees, and which explicitly showed that
those animals do possess a real organized way of communication. Consequently, the scientific
discoveries in this research field clearly confirm the main concept of animal communication
revealed in the holy Quran, and which was reported 14 centuries before.
• Effect of the holy Quran in Soul Appeasement and Treatment of Anxiety:
An experimental Evidence on the Divinity of the Book, we noticed a real positive effect of the
holy Quran on the appeasement of anxiety and the improvement of motivation. In fact, several
experimental investigations were made on real patients, and strangely, all experiments have
shown that listening to the recitation of the Quran does have a real positive impact on the soul
appraisement and anxiety reduction. But what could be the secret in such spiritual strength?
Sincerely, we do not see any possible cause except the fact that the holy Quran should be sent
down by the Creator (Allah, praise be upon Him) who embedded all the necessary treatment
required in such situations in His mysterious book.
•
Quran folios, Sanaa Quran folios and their corresponding verses in the Uthmanic compilation
of the present Quran, where the comparative analysis was based on characters and words. That
is, according to the radiocarbon dating of the ancient folios and the obtained statistical results,
we were able to confirm that the present holy book (at least in its skeletal morphology) should
represent an authentic copy of the first original Quran that was recited by the Prophet fourteen
centuries ago.
Finally, we can say that the discoveries reported in this works have shed light on an old enigma that had
not been solved (scientifically) for 14 centuries.
Again, the different investigations and results cited in this book can only reflect the authenticity, nobility,
supremacy, and divinity of the holy Quran.
Nevertheless, it is clear that further research in this domain is essential, since continued scientific
investigations will probably yield increasingly significant and impressive findings regarding this
extraordinary holy Book.
© Copyright 2021-2025 Sayoud.net. All Rights Reserved. 165
27 Personal Feeling
Many questions could interest the curiosity of the researcher inside his personal field/box of interest (in
a specific limited field), leading him to try making an effort to understand what is happening inside that
field/box of interest...
When he manages to understand some of the related theories (in that specific limited field), he may
of his box of interest, he may experience a strange feeling of sadness due to the ignorance about what
could exist outside that box. But since he decided to raise his eyes (towards the huge universe), that is
Thus, it has been important for me to try looking for the truth through the scientific analysis of what
does exist between our hands (i.e. the hoy Quran). Consequently, several investigations have been
conducted and analysed carefully to avoid misinterpretations or subjective assessments.
Also, by preparing this book (3rd edition), I tried to do my best to avoid any error or mistake, due to the
delicacy of the subject. However, I still present my sincere apologies in case of any unintentional mistake
found in this document.
Hopefully, may I ask this little book to continue its journey across the earth globe, illuminating the
minds and hearts of readers.
© Copyright 2021-2025 Sayoud.net. All Rights Reserved. 166
References
Abdelkafy. (2013). Last visit in. http://www.abdelkafy.com
Al-Batineh, M. (2019). Using Forensic Stylistics and Corpus Analysis for Detecting and
and Sir Richard Burton. New Voices in Translation Studies, 21.
al-Ghazali, M. (s. d.). From Wikipedia, the free encyclopedia (website). Last visit in 2021.
Al-Kaheel A. (s. d.). Al-Kaheel A. : Hearts Wherewith they Understand,
https://kaheel7.net/?p=2303&lang=en. Last visit on september 12, 2022.
Almujaiw, S. (2017). Grammatical construction of function words between old and modern
-based analysis. Corpus Linguistics and Linguistic Theory, 13(2).
Alqaradawi. (2013). Last visit in. https://www.al-qaradawi.net/.
Al-Qarni, A. (s. d.). Last visit in 2021. http://en.wikipedia.org/wiki/Aaidh_ibn_Abdullah_al-
Qarni.
Al-Rassi, M. S. (s. d.). Al-Rassi, M. S. The Amazing Prophecies OF Muhammad in the Bible,
2019. [Dawood, A. A.] David Abdu Benjamin Keldani, Muhammad in the Bible. Adam
Publishers 2007, page 189.
Al-Shreef. (2009). A. Al-Shreef. Is the Holy Quran Muhammad’s invention?
Amr-Khaled. (2013). Last visit in. http://amrkhaled.net
Ansorge, C. (2011). Ansorge, C., 2011. Faith & Fable : Islamic manuscripts from Cambridge
University Library. Cambridge University Publisher.
Anthony Matthew Jacob. (s. d.). Anthony Matthew Jacob, Is Prophet Muhammed (s.a.w.a.) In
the Bible ? Publisher World Islamic Network, 2017.
Armour, J. A. (s. d.). Armour, J. A. (2007). The little brain on the heart. Cleveland Clinic
Journal of Medicine, 74, S48-51.
Awwad K. (1982). Awwad K., 1982. Aqdam al-makhtutat al-arabiyya fi maktabat al-
alam (The Oldest Arabic Manuscripts in the World’s Libraries), Baghdad, 1982.
Ayad, A. (s. d.). Ayad, A. (2008). Healing body & soul : Your guide to holistic wellbeing
following Islamic teachings. International Islamic Publishing House.
BeesWiki. (s. d.). BeesWiki, 2023. How Do Bees Communicate ? BeesWiki,
https://beeswiki.com/how-do-bees-communicate/ Last access in ugust 2023.
© Copyright 2021-2025 Sayoud.net. All Rights Reserved. 167
Birdfact. (s. d.). Birdfact. How Do Birds Communicate ? (All Methods Explained), Birdfact
https://birdfact.com/articles/how-do-birds-communicate, Last updated : 3 August 2022.
Birmingham University. (2015). Birmingham University, 2015. Birmingham Quran
manuscript dated among the oldest in the world. July 2015.
Http://www.birmingham.ac.uk/news/latest/2015/07/quran-manuscript-22-07-15.aspx.
Birmingham University. (2016). Birmingham University, 2016. About the Birmingham
Quran, FAQs, www.birmingham.ac.uk/ facilities/cadbury/quran-manuscript/faqs.aspx. Last
access in july 2016.
Clement, R., & Sharp, D. (2003). Ngram and Bayesian Classification of Documents for Topic
and Authorship. Literary and Linguistic Computing, 18(4), 423-447,.
Contributors. (2015). Contributors of Wikipedia. Central limit theorem. Consultation on
August, 27. https://en.wikipedia.org/wiki/Central_limit_theorem.
Corney, M. (2003). Analysing E-Mail Text Authorship for Forensic Purposes [Master
Thesis,]. Queensland University of Technology.
Cox, I.J., Miller, M.L., Bloom, J.A, Kalker, T, Fridrich, J., & J. (2007). Digital Watermarking
and Steganography. Elsevier.
Darabinia M., Gorji A.M.H., Afzali M.A. (2017). Darabinia M., Gorji A.M.H., Afzali M.A.
The effect of the Quran recitation on mental health of the Iranian medical staff. Journal of
Nursing Education and Practice 2017, Vol. 7, No. 11. Http://jnep.sciedupress.com.
Dasarathy, B. V. (1994). Decision Fusion". IEEE Computer Society Press.
David Abdu Benjamin Keldani. (s. d.). David Abdu Benjamin Keldani, Muhammad in the
Bible. Adam Publishers 2007, page 189.
Denffer A.V. (2009). Denffer A.V., 2009. An Introduction to the Sciences of the Quran.
Publisher : Islamic Foundation (UK).
Digital Humanities 2010 Conference
Ellis, G., Mansmann, F., & VisMaster, V. A. (2010). Mastering the Information Age.
Chapter, 2. http://www.vismaster.eu/book/chapter-2-visual-analytics.
El-Zohairy, N. (2008). A Dictionary of Function Words in Arabic. Liban Publisher.
Evert, S. (2017). Understanding and explaining Delta measures for authorship attribution.
Digital Scholarship in the Humanities, 32(suppl_2), 4-16,.
Farringdon, J. M. (1996). Analyzing for Authorship : A Guide to the Cusum Technique.
University of Wales Press.
© Copyright 2021-2025 Sayoud.net. All Rights Reserved. 168
Fedeli A. (2015). Fedeli A., 2015. Phd thesis dedicated to the Qurʾan folios of Birmingham :
Alba Fedeli, Early Qur’ānic manuscripts, their text, and the Alphonse Mingana papers held
in the department of special collections of the University of Birmingham, 2015.
Ferrer-i-Cancho, R., & McCowan, B. (s. d.). Ferrer-i-Cancho, R., & McCowan, B. (2009). A
law of word meaning in dolphin whistle types. Entropy, 11(4), 688-701.
Foster, D. (2001). Author Unknown : On the Trail of Anonymous. Henry Holt and Company.
Fredrickson BL, Branigan C. (s. d.). Fredrickson BL, Branigan C (2005) “Positive emotions
broaden the scope of attention and thought-action repertoires.” Cognition and Emotion 19 :
313-332.
Fulton, J. T.,. (s. d.). Fulton, J. T., 2021. Deciphering the Dolphin Language Appendix U,
2021.
Galton, 1889] F. Galton. (1889). Book. Natural Inheritance, 66.
García-Barrero, D., Feria, M., & Turell, M. T. (2012). Using function words and punctuation
marks in Arabic forensic authorship attribution. IAFL Conference Proceedings.
Gavgani Z., V., Ghojazadeh, M., Sadeghi-Ghyassi, F., & Khodapanah, T. (2022). Gavgani Z.,
V., Ghojazadeh, M., Sadeghi-Ghyassi, F., & Khodapanah, T. (2022). Effects of listening to
Quran recitation on anxiety reduction in elective surgeries : A systematic review and meta-
analysis. Archive for the Psychology of Religion, 44(2), 111–126.
Https://doi.org/10.1177/00846724221102198.
Ghiasi A., Keramat A.,. (2018). Ghiasi A., Keramat A., The Effect of Listening to Holy Quran
Recitation on Anxiety : A Systematic Review, 2018, Iranian Journal of Nursing and Midwifery
Research | Published by Wolters Kluwer – Medknow.
Greenacre, M. (2014). Hierarchical cluster analysis, online documentation. Chapter, 7, 7 1-7
11.
H. Hadjadj, & Sayoud, H. (2018). Authorship Attribution of Seven Arabic Religious Books-A
Fusion Approach. Conference Paper: 6ème Conférence Internationale En Automatique &
Traitement de Signal.
Hart, T., Frank, D. D., Lopes, L. E., Olivos-Cisneros, L., Lacy, K. D., Trible, W., ... &
Kronauer, D. J. (s. d.). Hart, T., Frank, D. D., Lopes, L. E., Olivos-Cisneros, L., Lacy, K. D.,
Trible, W., ... & Kronauer, D. J. (2023). Sparse and stereotyped encoding implicates a core
glomerulus for ant alarm behavior. Cell, 186(14), 3079-3094.
HartMath. (s. d.). HartMath : The Heart–Brain Connection,
https://www.heartmath.com/science/. Last visit in august 2022.
© Copyright 2021-2025 Sayoud.net. All Rights Reserved. 169
Hawashin, B., Mansour, A., & Aljawarneh, S. (2013). An Efficient Feature Selection Method
International Journal of Computer Applications, 83-Number
17
Hickling R. (s. d.). Hickling R. Ants Have an Acoustic World of Their Own. Robert Hickling –
Exploresound, https://exploresound.org/2020/01/ants-have-an-acoustic-world-of-their-own/
Last access in August 2023.
Higham T. F., Bronk Ramsey G. C., Chivall D., Graystone J., Baker D., Henderson E.,
Ditchfield P. (2018). Higham T. F., Bronk Ramsey G. C., Chivall D., Graystone J., Baker D.,
Henderson E., Ditchfield P., 2018. Radiocarbon Dates from the Oxford AMS System,
Archaeometry Journal : Datelist 36. Vol. 60, Issue3, June 2018, Pages 628-640.
Holmes, D. (1998). The Evolution of Stylometry in Humanities Scholarship. Literary and
Linguistic Computing, 13(3),
Hopwood D. (1961). Hopwood D., 1961. ‘The Islamic Manuscripts in the Mingana
collection’, Journal of the Royal Asiatic Society XCIII/3-4.
Hrncir, M., Barth, F. G., & Tautz, J. (s. d.). Hrncir, M., Barth, F. G., & Tautz, J. (2005). 32
vibratory and airborne-sound signals in bee communication (hymenoptera). Insect sounds
and communication : Physiology, behaviour, ecology, and evolution, 421.
Huang, X., & Pan, W. (2003). Linear regression and two-class classification with gene
Bioinformatics, 19(16), 2072-2078,.
Huberty, C. J. (1994). Applied discriminant analysis. Wiley.
Ibnu-Kathir. (s. d.).
.
Ibrahim., I. 1997] I. A. (1997). A brief illustrated guide to understanding Islam. Library of
Congress, Catalog Card Number : 97-67654, Published by Darussalam, Publishers and
Distributors. http://www.islam-guide.com/contents-wide.htm.
Ibrahim, I. A. (1996). A brief illustrated guide to understanding Islam. Library of Congress,
Catalog Card Number : 97-67654, Published by Darussalam, Publishers and Distributors.
http://www.islam-guide.com/contents-wide.htm,
Illias S. (s. d.). Illias S., 2018. The Language of the Birds, Article date : September 19, 2018.
Signs & Science : The Language of The Birds (signsandscience.blogspot.com)
https://signsandscience.blogspot.com/2018/09/the-language-of-birds.html.
Islahi, I. 1989] A. A. (1989). Fundamentals of Hadith Interpretation – an English translat. Of
“Mabadi Tadabbur-i-Hadith (T. M. Hashmi, Trad.). Al-Mawrid.
© Copyright 2021-2025 Sayoud.net. All Rights Reserved. 170
Islam-beliefs. (s. d.). Islam-beliefs. Prophetic Lineage in Quran. Https://islam-
beliefs.net/en_US/prophetic-lineage-in-islam/. Last access in April 2023.
Jain, A. K., Ross, A., & Prabhakar, S. (2004). An Introduction to Biometric Recognition.
IEEE Transactions on Circuits and Systems for Video Technology, 14, number 1.
Jankowski, M., Hajjar, F., Kawas, S. A., Mukaddam-Daher, S., Hoffman, G., McCann, S. M.,
& Gutkowska, J. (s. d.). Jankowski, M., Hajjar, F., Kawas, S. A., Mukaddam-Daher, S.,
Hoffman, G., McCann, S. M., & Gutkowska, J. (1998). Rat heart : A site of oxytocin
production and action. Proceedings of the National Academy of Sciences,.
Juola, P. (2006). JGAAP: Authorship Attribution. Foundations and TrendsR in Information
Retrieval, 1(3), 233-334,.
Juola, P. (2009). JGAAP: A System for Comparative Evaluation of Authorship Attribution.
Proceedings of the Chicago Colloquium on Digital Humanities and Computer Science, 1(1).
Kaheel, A. (2015). The Marvels of number 7 in the noble Quran.
http://kaheel7.com/Book/Marvels_BookSeven.pdf.
Kalgutkar, V., Stakhanova, N., Cook, P., & Matyukhina, A. (2018). Android authorship
attribution through string analysis. Proceedings of the 13th International Conference on
Availability, Reliability and Security, 1-10 ,.
Keerthi, S. S., Shevade, S. K., Bhattacharyya, C., & K.R.K. (2001a). Murthy, Improvements
Neural Computation, 13, 637-649,.
Keerthi, S. S., Shevade, S. K., Bhattacharyya, C., & K.R.K. (2001b). Murthy, Improvements
Neural Computation, 13, 637-649,.
Kestemont, M., Manjavacas, E., Markov, I., Bevendorff, J., Wiegmann, M., Stamatatos, E., &
Stein, B. (2020). Overview of the Cross-Domain Authorship Verification Task at PAN 2020.
In CLEF.
Visualization. Conference: Genetic and Evolutionary Computation - GECCO 2003, Genetic
and Evolutionary Computation Conference
Kohonen, T. (1990). The self-organizing map. Proceedings of the IEEE, 78(ue: 9),
Kroopnick, Kroopnick, M. H., Chen, J., & Jaehwa Choi, C. M. D. (2010). Assessing
-Out
Methods. Journal of Modern Applied Statistical Methods. May, 9(1), 2-331 52-63.
Lachenbruch, P. A. (1967). An almost unbiased method of obtaining confidence interval for
the probability of misclassification in discriminant analysis
© Copyright 2021-2025 Sayoud.net. All Rights Reserved. 171
Lammers, M. O., Au, W. W., & Herzing, D. L. (s. d.). Lammers, M. O., Au, W. W., &
Herzing, D. L. (2003). The broadband social acoustic signaling behavior of spinner and
spotted dolphins. The Journal of the Acoustical Society of America, 114(3), 1629-1639.
Lazarus, R. S., & Folkman, S. (1984). Lazarus, R. S., & Folkman, S. (1984). Stress, appraisal,
and coping. Springer publishing company.
Li, J., Zheng, R., & Chen, H. (2006). From fingerprint to writeprint. Communications of the
ACM, 49
Love, H. (2002). Attributing Authorship : An Introduction. Cambridge University Press.
Madigan, D., Genkin, A., Lewis, D. D., Argamon, S., Fradkin, D., & Ye, L. (2005). Author
identification on the large scale. Joint Annual Meeting of the Interface and the Classification
Society of North America (CSNA.
McCowan, B., Hanser, S. F., & Doyle, L. R. (s. d.). McCowan, B., Hanser, S. F., & Doyle, L.
R. (1999). Quantitative tools for comparing animal communication systems : Information
theory applied to bottlenose dolphin whistle repertoires. Animal behaviour, 57(2), 409-419.
McCraty, R. (s. d.). McCraty, R. (2015). Science of the heart : Exploring the role of the heart
in human performance. HeartMath Research Center, Institute of HeartMath.
McCraty R, Atkinson M, Tomasino D, Bradley RT. (s. d.). McCraty R, Atkinson M, Tomasino
D, Bradley RT. The coherent heart : Heart–brain interactions, psychophysiological
coherence, and the emergence of system-wide order.
McLachlan 2003] G.J. McLachlan, R. W. Bean., D. Peel. (2003). Modelling high-dimensional
data by mixtures of factor analyzers. Computational Statistics & Data Analysis, 41(ues 34),
McLnchlan 2001] G. J. McLachlan, D. Peel., S. K. Ng. (2001). On Clustering by Mixture
Models,Proceedings of the 25th Annual Conference of the Gesellschaft für Klassifikation e.V.
University of Munich.
McMenamin, G. R. (2002). Forensic Linguistics—Advances in Forensic Stylistics. CRC
Press.
Mercer D. (s. d.). Mercer D. 2022, Can I talk to my Hawk ? | Bird of Prey Communication—
Mercer Falconry, https://www.youtube.com/watch?v=9mWzwWY1TuM, oct. 2022.
Mills, D. E. (2003). Authorship Attribution Applied to the Bible [[Mills 2003]]. Graduate
Faculty of Texas, Tech University.
Milne, & Milne, W. E. (2012). Numerical Calculus—Approximations, Interpolation, Finite
Differences, Numerical Integration, and Curve Fitting.
© Copyright 2021-2025 Sayoud.net. All Rights Reserved. 172
Mohamed-Cherif, M. I. (2007). Interogative styles in rethorical research and its secrets in the
Quran [PhD thesis,]. International Islamic University.
Int. Conference on Numerical Inimitability in
Quran 7-9 Nov. 2008.
Mosteller, F., & Wallace, D. L. (1964). Inference and Disputed Authorship : The. Addison-
Wesley.
Muangprathub, J., Kajornkasirat, S., & Wanichsombat, A. (2021). Document Plagiarism
Detection Using a New Concept Similarity in Formal Concept Analysis. Journal of Applied
Mathematics.
Muqith M. A. (2011). Muqith M. A., 2011 Al-Madd Al-Tabee’ee. Heesbees : All for Quran
and Tajweed. January 2011. Https://heesbees.wordpress.com/tag/short-and-long-arabic-
vowels/.
Naghich, N. (2012). Interogative style in the Prophetic statements. MSc Thesis.
Nasr, S. H. (2004). The heart of Islam : Enduring values for humanity (Harper, SanFrancisco,
c2002).
Norusis 2008], M. Norusis. (2008). Cluster Analysis, Chapter 16.,pp:361-391. SPSS 17.0
Statistical Procedures Companion. Marija Norusis.
Ouamour, Ouamour, S., & Sayoud, H. (2013). Authorship Attribution of Short Historical
Arabic Texts Based on Lexical Features. CyberC – International Conference on Cyber-
Enabled Distributed Computing and Knowledge Discovery CyberC Conference.
Peng, Peng, F., Schurmans, D., Keselj, V., & Wang, S. (2003). Language independent
authorship attribution using character level language models. Proceedings of the 10th
Conference of the European Chapter of the Association for Computational Linguistics, 267-
274,.
Powers D. S. (2011). Powers D. S., 2011. Review : La transmission écrite du Coran dans les
débuts de l’islam : Le codex Parisino-petropolitanus. Texts and Studies on the Qurʾān, 5 by
François Déroche. Islamic Law and Society, Vol. 18, No. 2 (2011), pp. 281-285.
Quran Project. (s. d.). Old and New Testament Prophecies of Muhammad,
https://www.quranproject.org/Quran-Project-Appendix-Old-and-New-Testament-Prophecies-
of-Muhammad—227-d, Last access in April 2023.
Ramsey B. C., Higham T. F. G., Brock F., Baker D., & Ditchfield P. (2009). Ramsey B. C.,
Higham T. F. G., Brock F., Baker D., & Ditchfield P., (2009). Radiocarbon dates from the
Oxford AMS system : Archaeometry datelist 33. Archaeometry, 51(2), 323-349.
Archaeometry, 51(2), 323-349.
© Copyright 2021-2025 Sayoud.net. All Rights Reserved. 173
Rogers L. J. & Kaplan G. (s. d.). Rogers L. J. & Kaplan G.. (2002). Songs, roars, and rituals :
Communication in birds, mammals, and other animals. Harvard University Press, 2002.
Sadeghi, B., & Bergmann, U. (2010). Sadeghi, B., & Bergmann, U. (2010). The Codex of a
Companion of the Prophet and the Qurān of the Prophet. Arabica, 57(4), 343-436.
Saged, A.A.G., Mohd Yusoff, M.Y.Z., Abdul Latif, F. et al. (2020). Saged, A.A.G., Mohd
Yusoff, M.Y.Z., Abdul Latif, F. et al. Impact of Quran in Treatment of the Psychological
Disorder and Spiritual Illness. J Relig Health 59, 1824–1837 (2020).
Https://doi.org/10.1007/s10943-018-0572-8.
Sanaa. (1985). Sanaa, 1985. Maṣāḥif Ṣanʿāʾ, 1985, Dār al-Athar al-Islamiyyah : Kuwait, p.
59, Plate 4.
Sanderson, C., & Guener, S. (2006). Short text authorship attribution via sequence kernels,
Proceedings of International
Conference on Empirical Methods in Natural Language Processing (EMNLP
features for authorship attribution. Proceedings of the 27th International Conference on
Computational Linguistics, 343-353 ,.
Sarwar, R., Urailertprasert, N., Vannaboot, N., Yu, C., Rakthanmanon, T., Chuangsuwanich,
-Author
Documents Using a Co-Authorship Graph. IEEE Access, 8, 18374-18393,.
Sayoud, H. (2003). Automatic speaker recognition – Connexionnist approach” [PhD thesis,].
USTHB University.
LLC Journal, Literary and Linguistic Compting, 27(4), 427-444,.
Sayoud, S. 2010] H. (2010). Investigation of Author Discrimination between two Holy
Islamic Books. IET (Ex-IEE) Teknologia Journal, 1(ue. 1), ,.
statements. Literary and Linguistic Computing, 27
https://doi.org/10.1093/llc/fqs014
Schönrogge, K., Barbero, F., Casacci, L. P., Settele, J., & Thomas, J. A. (s. d.). Schönrogge,
K., Barbero, F., Casacci, L. P., Settele, J., & Thomas, J. A. (2017). Acoustic communication
within ant societies and its mimicry by mutualistic and socially parasitic myrmecophiles.
Animal Behaviour, 134, 249-256.
Schuster, T., Schuster, R., Shah, D. J., & Barzilay, R. (2020). The limitations of stylometry
for detecting machine-generated fake news. Computational Linguistics, 46(2), 499-510,.
© Copyright 2021-2025 Sayoud.net. All Rights Reserved. 174
Shlens, J. (2003). A TUTORIAL ON PRINCIPAL COMPONENT ANALYSIS - Derivation,
Discussion and Singular Value Decomposition. J. Shlens, 1.
Signoriello, D. J., Jain, S., Berryman, M. J., & Abbott, D. (2005). Advanced text authorship
detection methods and their application to biblical texts. Proceedings of SPIE, 6039
Townsend, S. W. (s. d.). Spiess, S., Mylne, H. K., Engesser, S., Mine, J. G., O’Neill, L. G.,
Russell, A. F., & Townsend, S. W. (2022). Syntax-like Structures in Maternal Contact Calls of
Chestnut-Crowned Babblers (Pomatostomus ruficeps). International Journal of Primatology,
1-20.
Stamatatos, E. (2007). Author identification using imbalanced and limited training texts.
Proceedings of the 4th International Workshop on Text-Based Information Retrieval,
Stamatatos, E. (2009). A survey of modern authorship attribution methods. Journal of the
American Society for Information Science & Technology, 60
Stamatatos, Stamatatos, E., Fakotakis, N., & Kokkinakis, G. (2001). Computer-based
authorship attribution without lexical measures. Computers and the Humanities, 35(2), 193-
214,.
Stylianou, Y., & et al. (2005). Y. Stylianou, Y. Pantazis, F. Calderero, P. Larroy, F. Severin,
S. Schimke, R. Bonal, F. Matta, , and A. Valsamakis,“GMM- Based Multimodal Biometric
Verification,” Final Project Report 1, Enterface’05, July 18th—August 12th, Mons, Belgium,
2005.
Suganya, S. 2012] R., & R. Shanthi, F. C.-M. A.-. (2012). A Review, International Journal of
Scientific and Research Publications, 2(ue 11).
T. Kardi, K-Means Clustering Tutorial. E-Book by Teknomo. (2007). Kardi.
http://people.revoledu.com/kardi/tutorial/kMean/index.html
Tambouratzis, G. (2004). Discriminating the Registers and Styles in the Modern Greek
Language-Literary and Linguistic Computing, 19(2),
197-220.
Tambouratzis, G., Markantonatou, S., Hairetakis, N., Vassiliou, M., Tambouratzis, D., &
Carayannis, G. (2000). Discriminating the Registers and Styles in the Modern Greek
Language. In Proceedings of the Workshop on Comparing Corpora (held in conjunction with
the 38th ACL Meeting
Tambouratzis, G., Markantonatou, S., Vassiliou, M., & Tambouratzis, D. (2003). Employing
Statistical Methods for Obtaining Discriminant Style Markers within a Specific Register. In
Proceedings of the Workshop on Text Processing for Modern Greek : From Symbolic to
© Copyright 2021-2025 Sayoud.net. All Rights Reserved. 175
Statistical Approaches (held in conjuction with the 6th International Conference of Greek
Linguistics), Rethymno
Taylor R.E. and Aitken M.J. (1997). Taylor R.E. and Aitken M.J. 1997. Chronometric dating
in Archaeology. Advances in Archaeological and Museum Science, volume 2. Oxford
University, England.
Tchechmedjiev, A., Schwab, D., & Goulian, J. (2013, mars 24). Fusion strategies applied to
multilingual features for an knowledge-
CICLING’ 2013 Conference: 14th International Conference on
Intelligent Text Processing and Computational Linguistics.
UNESCO. (1980). UNESCO, 1980. « Recommendations » in R. B. Serjeant (Ed), The Islamic
City, 1980, UNESCO: France, pp. 207-208.
Verlinde, P. (1999). A Contribution to Multimodal Identity Verification using Decision Fusion
[PhD thesis,]. Ecole Nationale Supérieure des Télécommunications.
Von Frisch, K. (s. d.-a). Von Frisch, K. (1965). Die Orientierung der Bienen unterwegs zum
Ziel. Tanzsprache und Orientierung der Bienen, 331-537.
Von Frisch, K. (s. d.-b). Von Frisch, K. (1965). Die tänze der bienen (pp. 3-330). Springer
Berlin Heidelberg.
Whiteman, & Whiteman, M. (1967). Philosophy of Space and Time and the Inner
Constitution of Nature : A Phenomenological StudyRelié– 1967. Humanities Press.
Wiki1. (2012). Quran. The free encyclopedia. Wikipedia, Last modified in.
http://en.wikipedia.org/wiki/
Wiki2. (2012). Hadith. The free encyclopedia. Wikipedia, Last modified in.
http://en.wikipedia.org/wiki/
Wiki_COS. (2013). Wikipedia, Cosine similarity, From Wikipedia, the free encyclopedia. The
web page was last modified on. http://en.wikipedia.org/wiki/Cosine_similarity.
Wiki_REG. (2013). Linear regression, From Wikipedia, the free encyclopedia. The web page
was last modified on. http://en.wikipedia.org/wiki/Linear_regression.
Practical machine learning tools and techniques with Java implementations. In N. Kasabov &
K. Ko (Éds.), Proceedings of the ICONIP/ANZIIS/ANNES’99 Workshop on Emerging
Knowledge Engineering and Connectionist-Based Information Systems (p. 192-196,).
Zangerle, E., Mayerl, M., Specht, G., Potthast, M., & Stein, B. (2020). Overview of the style
change detection task at PAN 2020. CLEF.
© Copyright 2021-2025 Sayoud.net. All Rights Reserved. 176
Zouhir, A., El Ayachi, R., & Biniz, M. (2021). A comparative Plagiarism Detection System
methods between sentences. Journal of Physics, 1743(1), 012041.
© Copyright 2021-2025 Sayoud.net. All Rights Reserved. 177
Note: This book is made under the OPEN ACCESS rules and policies, with no commercial use.
The book is published under the terms of a Creative Commons license, which permits use and
distribution, but no commercial use.
Book Edited by Pr Halim Sayoud
Publisher EDT - SCHOLARPAGE
© Copyright Sayoud.net, 2021-2025, All rights reserved
DOI: 10.5281/zenodo.14865663
February 2025
