A Comparative Study of Computational Fake News Detection on Social Media

Abstract

Social media platforms has become widely popular among netizens to share views and news. This can be attributed to easy affordability of digital devices, low cost Internet and free subscription to social media platforms. Individuals find social media platforms quite appealing where they can find like-minded people to exchange views and news. Studies have shown that a less credible person is more likely to propagate fake news in order to fulfill their objectives of any form - seeking attention, gaining financial benefits or influencing political views. Hence, fake news detection on social media has become one of the most envisaged research topics in recent times. Fake news detection on social media can be done through various methods based on sources, transmission, styles, and knowledge. The major contribution of this survey paper is to provide a comparative study of major computational methods for fake news detection on social media. The objective of this study is to help researchers carry further research using the details and the research gaps presented in the paper.

Full text

A Comparative Study of Computational Fake News
Detection on Social Media
Manish Kumar Singh Manish (  manish.kumar2191989@gmail.com )
Jamia Hamdard
Jawed Ahmed Jawed
Jamia Hamdard
Mohammad Afshar Alam Alam
Jamia Hamdard
Kamlesh Kumar Raghuvanshi Kamlesh
University of Delhi
Sachin Kumar Sachin
University of Delhi
Research Article
Keywords: Social Media Hoax, Fake News Detection, Fake News Dataset, PRISMA, Systematic Review
Posted Date: August 18th, 2022
DOI: https://doi.org/10.21203/rs.3.rs-1964791/v1
License:   This work is licensed under a Creative Commons Attribution 4.0 International License. 
Read Full License

A Comparative Study of Computational
Fake News Detection on Social Media
Manish Kumar Singh1*, Jawed Ahmed1†, Mohammad Afshar
Alam1†, Kamlesh Kumar Raghuvanshi2†and Sachin Kumar3†
1Department of Computer Science, Jamia Hamdard, Hamdard
Nagar, New Delhi, 110062, Delhi, India.
2Department of Computer Science, Ramanujan College,
University of Delhi, Kalkaji, New Delhi, 110019, Delhi, India.
3Cluster Innovation Center, University of Delhi, GC Narang
Road, New Delhi, 110007, Delhi, India.
*Corresponding author(s). E-mail(s):
manish.kumar2191989@gmail.com;
Contributing authors: jahmed2047@jamiahamdard.ac.in;
aalam@jamiahamdard.ac.in;kamlesh@ramanujan.du.ac.in;
skumar@cic.du.ac.in;
†These authors contributed equally to this work.
Abstract
Social media platforms has become widely popular among netizens to
share views and news. This can be attributed to easy affordability of dig-
ital devices, low cost Internet and free subscription to social media plat-
forms. Individuals find social media platforms quite appealing where they
can find like-minded people to exchange views and news. Studies have
shown that a less credible person is more likely to propagate fake news
in order to fulfill their objectives of any form - seeking attention, gaining
financial benefits or influencing political views. Hence, fake news detec-
tion on social media has become one of the most envisaged research topics
in recent times. Fake news detection on social media can be done through
various methods based on sources, transmission, styles, and knowledge.
The major contribution of this survey paper is to provide a comparative
study of major computational methods for fake news detection on social
media. The objective of this study is to help researchers carry further
research using the details and the research gaps presented in the paper.
1

2A Comparative Study of Computational Fake News Detection on Social Media
Keywords: Social Media Hoax, Fake News Detection, Fake News Dataset,
PRISMA, Systematic Review
1 Introduction
Social media platforms has become a kind of democratized media among neti-
zens to share views and news at high frequency these days [1]. This is evident
in developing countries where 62 percent users retrieve news from social media
platforms [2]. This can be attributed to advancing technology, easy affordabil-
ity of digital devices, low cost Internet, ease of access and free subscription
to social media platforms [3]. As a result, the traditional news media such as
television, newspapers, radio, magazines etc. are transitioning to a digital form
such as social media platforms (Facebook, Twitter, Youtube, etc.), blogs, news
sites etc. in order to reach large number of people in various manner [4]. Com-
pared to the other digital medias discussed above, social media platforms has
the ability to connect large audiences than traditional media. Further social
media, if handled ideally, proves to be a powerful tool for both businesses
and individuals in order to exchange and spread information in a shorter time
than ever before. However, it is evident that the information obtained by con-
sumers through social media platforms is not necessarily true and may not
reach them promptly. To make the things worse, these platforms are being
actively used to spread false information, rightly termed as ”fake news” [5], in
order to influence opinion of others for mere personal gains [4].
There are two main motivations behind the mushrooming of fake news
related sites and social media posts [2]: a). pecuniary related intentions, that
involve drawing significant amount of revenue through viral news pieces; and
b). ideological related intentions, that revolve around influencing opinion of
people on specific subjects. Additionally, the rise of bots and trolls like online
malicious agents are the key contributors to the dissemination of the fake news
[6].
The term ”fake news” implies to the news that is purposely and verifiably
untrue [5]. It may contain entirely incorrect information or supposed to be
purely misinterpretation of a true event. Social media platforms and instant
messaging apps have made it quite easy to disseminate misleading information
to large audiences [7]. Fake news, due to its appealing nature, reaches among
large number of people quickly and has deep impact on people’s perceptions
related to numerous subject. This can lead to genocides and mass protests
that may further result into social, political and economic crises [8]. Last but
not the least, false news may have a big impact on real world. For instance, a
genuine shooting was sparked by the fake news story ”Pizzagate” from Reddit
[9]. Therefore, it is crucial to address the problem of detecting fake news.
Fake news is not a new phenomena [10]. Fake news and hoaxes have existed
long before the advent of the Internet [11]. It is especially pertinent at this
moment to ask why it has become a worldwide topic of concern and why it

A Comparative Study of Computational Fake News Detection on Social Media 3
is drawing rising public attention. The main reason for this is that the online
creation and publish of fake news are more faster and less expansive than tra-
ditional news media like television and newspapers [5]. The issue of fake news
on social media or say social media hoax has become a serious worry among
concerned stakeholders. It is quite obvious that an ordinary person may find it
difficult to verify the verasity of fake news due to lack of expertise. Hence, the
research in the area of building reliable and effective fake news detection on
social media has caught the attention of academicians, researchers and indus-
try in the recent past. Consequently, many attempts have been made to apply
artificial intelligence (AI) based techniques to build automated and auxiliary
methods to enable early spotting of fake news on social media automatically
[12]. In short, the real objective of fake news detection on an automated basis
is to limit the time and effort required by people to identify fake news and
preventing it from spreading. To detect fake news dissemination on social
media platforms, several researches have been conducted using content-based
[13][14][15][16], social context-based [12][17] and propagation structure-based
[18][19] detection techniques regardless of their own limitations.
The following are the primary contributions of this paper:
1. it provides an organized and comprehensive theoretical background of fake
news detection research on social media;
2. it systematically analyzes and discusses the comparison among major exist-
ing works related to computational methods for detecting fake news on
social media that appeared during the period between the year 2011 and
April 2022;
3. it presents research gaps for further research in the area of fake news
detection on social media.
The remainder of the paper is structured as follows: section 2 provides the
theoretical background of fake news detection research on social media; section
3 presents the research methodology followed in carrying out the current work;
section 4 gives the comparative analysis of computational method based work
done in the area of fake news detection on social media between the year 2011
and April 2022; section 5 discusses our major research findings; sections 6
outline the research gaps for carrying future research in the fake news detection
area on social media; and section 7 finally concludes the paper.
2 Theoretical Foundation of Fake News
Detection Research on Social Media
2.1 What is Fake News?
The term ”fake news” has a long history. During the 2016 US presidential
election campaign, it earned a poor reputation and became a popular word [20].

4A Comparative Study of Computational Fake News Detection on Social Media
Fake news research gained momentum since then and it is now regarded as the
contamination of information with false news, propaganda, hoaxes, junk news
and rumours [21]. Even though these terms imply deceiving information, there
is currently no consensus on the definition or categorization of fake news, even
in journalism. Hence, a precise and accurate definition is required to better
do examination of fake news and assessment of pertinent studies related to it.
Hence, we have covered various popular fake news definitions in this section
as suggested by certain studies related to fake news detection on social media
-1. the definition of related concepts of fake news ;2. the broad definition of
fake News [22]; and 3. the narrow definition of fake news [5].
2.1.1 Definition of Related Concepts of Fake News
Below are the popular related terms of fake news which have appeared in many
recent published works:
•False News: It is defined as fabricated news stories that may be poten-
tially or purposefully misleading to readers since they resemble regular news
content in form but not in intent or organisational process [21][7].
•Propaganda: It implies to news stories that are designed to influence target
audiences’ emotions, views, and actions through deceit, selective omission of
or provision of one–sided content for ideological, political or religious goals
[10].
•Hoax: It refers to a lie created with the goal of passing itself off as the truth
[23].
•Rumor: Some academicians have classified it as a subcategory of propa-
ganda [24]. It comes from a Latin term which means ”noise.” It might spread
from one person to another as an unsubstantiated assertion that wasn’t the
result of news events [25].
•Satire: It is a type of news created with the intent of amusing or criticising
readers, and which may appear to be true news; this may be dangerous when
shared out of context [10][26].
•Junk News: It is a broader concept which typically refers to the entire
news content that applies to a publication instead of just one article and it
collects numerous forms of information [27].
•Click-bait: It is considered as low-quality journalism aimed at attracting
visitors and generating revenue through advertising [28][29].
All above related concepts of fake news fall under three categories:
•Misinformation: It can be simply defined as false information communi-
cated without an agenda by persons who are either unaware with the facts
or are cognitively handicapped[30][31].
•Disinformation: It comprises of the intentional, deliberate, or purposeful
dissemination, statement, or propagation of false, incorrect, or misleading
information with an agenda to mislead, deceive, or confuse [32][33].

A Comparative Study of Computational Fake News Detection on Social Media 5
•Mal-information: It is defined as a fact-based statement of information
but is employed to hurt an individual, a group, or a nation [34].
Table-1 briefly compares the related terms of fake news on the basis of
misinformation, disinformation, and mal-information.
Table 1: Comparison of Related Concepts of Fake News
Related Concept Misinformation Disinformation Mal-information
False News Yes Yes Yes
Propaganda No Yes Yes
Hoax Yes Yes No
Rumor Yes Yes No
Satire No No Yes
Junk News No Yes Yes
Click-bait No Yes No
Source: This table of comparison is based on the idea presented by
Elhadad et al.[35].
2.1.2 Fake News: A Broad Definition
The conventional ways of defining news have been challenged by the growing
digitalization of news. Nowadays, online platforms allow non-journalists to
access a sizable audience [10]. For these situations, Zhou and Zafarani [22]
define false news broadly as:
”Fake news is false news”
This broad definition of fake news broadly covers speeches, statements, claims,
articles, posts along with other information about a particular topic. Further,
both journalists and non-journalists are capable of creating such news. The
broad definition of fake news seeks to impose the fewest limits possible in light
of available resources. It focuses upon the veracity of information and uses a
broad meaning of the term ”news” with having some specific purpose [7] where
getting the real story is tough which simply reduces the need for information
related goals (i.e. true intention).
2.1.3 Fake News: A Narrow Definition
The more specific way of defining fake news that meets the overall criteria for
any news to be fake is given by Shu et al. [5] as:
”Fake news is a news article that is intentionally created with the goal to
deceive and that can be checked upon as false”
This definition covers the public’s view of fake news, particularly in the after-
math of the United States presidential election of 2016. The aforementioned

6A Comparative Study of Computational Fake News Detection on Social Media
narrow definition emphasises on both news authenticity and goals, while mak-
ing sure that the published material is news by establishing whether or not
the publisher is a source of news [22] (e.g., The New York Times, The Times
of India, India Today, etc.). News outlets frequently publish news as articles
with pre-determined elements including a heading, author(s), body of text,
picture(s), or video (s) containing assertions made by or about prominent
individuals and organisations.
2.2 Effects of Fake News on Real World
Fig. 1: A Study on Misinformation in India during Janata Curfew (Sources:
A Research on Misinformation from the University of Michigan, published on
April 18, 2020).
The sudden spike in the attention towards fake news can be explained con-
sidering a number of dramatic events that have recently occurred around the
globe. In fact, most of the cases related to fake news belong to politics, as
pointed by Vosoughi et. al [7]. The term ”fake news” has gained widespread
acceptance after the US presidential elections in 2016; and some claim that
Donald Trump might not have won the presidency if it weren’t for the effects
of false news (in addition to the suspected Russian troll interference) [2]. Fur-
ther, recent research has indicated that fake news affected the referendum of

A Comparative Study of Computational Fake News Detection on Social Media 7
UK Brexit in 2016 [36] besides the French presidential elections of 2017 [37].
Another series of notable fake news incidents witnessed at the global level
includes the stock market meltdown brought on by a fake tweet about Presi-
dent Obama in the year 2013 [38], the misleading news surrounding Pizzagate
led to a shooting incident in a restaurant [5] and the widespread scepticism
of vaccines that developed during the epidemics of Zika and Ebola virus [39].
The three-month-long diplomatic standoff in the Gulf in 2017 was sparked in
part by the release of a fake news report planted on a news agency in Qatar
(Middle East) by hackers [40]. In the Middle East, the dissemination of disin-
formation frequently results in a shift in ideology [41]. Almost two weeks after
Russia invaded Ukraine, various false reports circulated alleging the war was a
hoax, a media invention, or that the scope of the conflict had been exaggerated
by the West [42].
A research on misinformation in India found an increase in the amount of
disproved tales, particularly the Prime Minister Narendra Modi’s proclama-
tion of janata curfew on March 22, 2020, resulted into a nationwide lockdown
two days later in order to restrict the COVID-19 spread [43]. The Indian gov-
ernment ordered the blocking of 22 YouTube-based news channels for allegedly
spreading fake news to mislead viewers [44]. The number of real time effects
of fake news are by far high which is not possible to cover entirely here.
2.3 Fake News Detection Challenges
Fig. 2: Five V’s of Fake News as Five challenges for it’s detection.

8A Comparative Study of Computational Fake News Detection on Social Media
Zhang et al. [45] explained three Vs (Volume, Veracity and Velocity) as
three characteristics of fake news. After deep review on fake news detection
on social media platforms, we realised that two more Vs (Variety and Valid
Dataset) along with the three Vs as explained by Zhang et. al. [45] together
can be thought as five serious challenges for building an effective and efficient
fake news detection system:
•Veracity: Fake news is purposefully created to confuse readers and imitate
traditional news outlets, leading to an adversarial situation in which it is
difficult to tell the difference between true and false news [5][6][45].
•Volume: Because of the speed and volume with which fake news is pro-
duced, it is no longer possible to thoroughly fact-check and verify all things,
such as engaging human experts for verification of news articles [5]. This
also raises concerns about building tools for detecting fake news on early
basis in order to prevent it from spreading over the internet [46].
•Variety: Fake news can be defined in various interrelated ways such as
rumors, satire news, disinformation, fake reviews, propaganda, hoax, click-
bait, and so on, that affects myriad aspects of people’s life [45].
•Velocity: Fake news creators have a short lifespan [2]. The fact that fake
news is spread in real time on social media makes spotting it even more
difficult. It’s difficult to estimate how many online users are engaged with a
particular piece of viral news [45].
•Valid Dataset: Due to the constraints imposed by social media platforms
on the acquisition of public data, the research community has provided
extremely restricted training datasets for fake news detection, which often
do not include all of the information related to false news [47]. The valid
datastet is very much needed to check the veracity of a fake news.
2.4 How Can Fake News be Detected?
The detection of fake news on social media requires the basic understanding of
its definition on the basis of mathematical modelling, its data mining frame-
work, its model formulation and the possible dataset for training and testing
the model in order to detect whether a given news is fake or not. This sim-
ply implies that the detection of false news belongs to the binary classification
problem of machine learning [48]. In this section, we provide the mathemat-
ical modelling of fake news on social media, its data mining framework and
its detection architecture. While in the section-4.1, we have compared and
explained the existing fake news detection datasets.
2.4.1 Mathematical Modelling of Fake News Detection
A mathematical modelling of fake news detection can be given as: ”the purpose
of fake news detection ’F’ is to assess whether or not the news piece ’a’ is fake
based on the social news engagements ’E’ among ’n’ people” i.e. F : E {0,1}”
where F(a)=1 (if ’a’ is a fake news);
or F(a)=0 (if ’a’ is not a fake news).

A Comparative Study of Computational Fake News Detection on Social Media 9
This narrow definition is given by Shu et al. [5] which is quite promising not
only to understand the problem of fake new detection but it further helps to
work on the development of its framework as well as its model for checking
whether a given news is fake or not. This definition is in accordance with the
philosophy of publisher-induced distortion bias on information [48] which is
basically a binary classification problem.
2.4.2 Fake News Detection Framework
A data mining framework for detecting fake news is broadly consist of two
phases [5]: (A). the extraction of features; followed by (B). the creation of
model. Many works followed this framework to detect fake news on social
media including [49][50][12][17][51][52][13] and many more. These two phases
are explained below.
(A). Extraction of Features: The goal of the feature extraction phase
is to create a formal mathematical structure that represents news content and
related auxiliary data. Elhadad et. al. [35] categorically explains the three
popular ways of extracting features from social media for fake news detection
((a). news content, (b). social context, (c). domain-specific) which has been
displayed in a refined form in Fig.3.
a). News Content: The meta information about a piece of news is described
by news content features. Different kinds of feature representations may be
created based on the raw content attributes (like source of a news article, its
heading, its body text and embedded image/video) of news content to derive
discriminative aspects of fake news. The news items that is often used as a
part of feature extraction can be either linguistic or visual in nature.
•Linguistic based feature extraction: Fake news frequently employs opinion-
ated and inciting language, like ”clickbait” (i.e., to get people to click on
the link and read the full story), in order to confuse or gain financial/po-
litical advantage since fake news is published for these purposes rather
than to report on factual claims. To detect fake news, it is appropriate to
use linguistic based features that capture diverse writing styles and sen-
sational headlines [29]. The extraction of the linguistic based features can
be done from the text content either in the form of (i). lexical features
(consisting of features at the character and word levels, such as charac-
ters per word, distinctive words, total words, and frequency of large words),
as well as (ii). syntactic features (consisting of elements at the sen-
tence level, such as words and phrases with functions used frequently in the
form of ”n-grams”, BoW (bag-of-words) [53], or punctuation, POS (parts-of-
speech) tagging that are useful for representing documents in NLP (Natural
Language Processing) for several purposes.
•Visual based feature extraction: To capture the many characteristics of fake
news, the extraction of visual based feature can be done from visual materials

10 A Comparative Study of Computational Fake News Detection on Social Media
such as images and videos. For news verification, numerous visual features
(containing coherence score, clarity score, clustering score, diversity score,
and similarity distribution histogram) and statistical features (containing
count, image ratio and its associated concepts) can be retrieved [54].
.
b). Social Context: The process of news dissemination through time can
be easily represented by social media engagements, which act as a source of
auxiliary information for inferring the truthfulness of news pieces. Users and
posts are the two key characteristics of the social media context that may be
extracted to help fake news detection.
•User based feature extraction: Here, the profiles and features of users are
extracted which may exist either at individual level (in the form of number
of posts published by the user, number of followers or followees etc. that
is helpful to determine the credibility and reliability of the user [12]) or
at group level (in the form of the ’average number of followers’ and the
’proportion of verified people’ [55][56] that help to enumerate the general
features of groups of users who are interested in a particular news).
•Post based feature extraction: Here, the reactions or opinions of users towards
a social media post are extracted for fake news detection which may exist at
post level (in the form of feature values for each post that may be extracted
through the linguistic-based approach as discussed earlier as well as various
embedding techniques [57] for the given post on the basis of stance [58], topic
[56] and credibility [12]), group level (in the form of the aggregate feature
value of all significant posts i.e. average credibility score for particular news
pieces using the intelligence of the crowd [58]) and temporal level (in the
form of feature values on the basis of their temporal variations [56]).
c). Domain Specific: These features are matched to news of specific domains
and may contain external links, quoted words, the total number of graphs and
their average length among other things [59]. These domain specific features
can be extracted either through network based or propagation based techniques.
•Network based feature extraction: Here, the features of different networks
(that are formed on social media platforms on the basis of topic of interest)
like stance network [58][60], co-occurrence network [57], diffusion network
[55] etc. are extracted by building specific network among users according
to the post published by them on a specific social media platform.
•Propagation based feature extraction: The fake news have propagation pat-
terns different from that of the real news [61] which is a clear indication
that propagation features can be used for classifying a news to be fake or
not. Nevertheless, a very few works have used these features for the job of
detecting fake news so far.
(B). Creation of Model: During the model creation phase, machine
learning models are built to better distinguish false news from actual news
based on the information obtained from the extracted features. The existing

A Comparative Study of Computational Fake News Detection on Social Media 11
Fig. 3: Extracted Features Types in Fake News Detection Framework.

12 A Comparative Study of Computational Fake News Detection on Social Media
works have used two kinds of fake news detection model on the basis of what
type of feature extracted from related published news on social media platform
- a). news content based model on the basis of news content based extracted
features, and b). social context based model on the basis of social context based
extracted features.
a). News Content Based Model: The existing approaches for constructing
news content based model of fake news detection may be either knowledge
based approach or style based approach.
•Knowledge based approach: This approach seeks to fact-check proposed
statements in news articles using external sources in order to check the
veracity of claimed statement in related news article in a certain situation
[62]. There are three fact checking approaches that have been utilised so
far for building knowledge-based fake news detection model - (i). expert-
oriented fact-checking (it employs human experts of particular domains
to refer valuable data/documents in order to verify the truthfulness of the
claimed statement in the news even though this approach requires highly
skilled manpower and lots of time), (ii). crowd-sourced fact checking (it
employs large population of ordinary people to annotate given news followed
by the overall evaluation of each annotation in to order to check the truth-
fulness of the claim made in the news even though this approach is difficult
to handle and is less accurate as well as less credible), and (iii). auto-
matic fact-checking (it employs the techniques of Information Retrieval,
Natural Language Processing, Machine Learning and graph theory to build
a automatic scalable system to verify the truthfulness of the given news’
claimed statement [63]). Table-2 provides the detailed examples of the above
discussed fact-checking approaches.
•Style based approach: This approach aims to detect fake news by locat-
ing the manipulators in the news content writing style. There are two
approaches that have been employed so far for building style-based fake news
detection model - (i). deception-based approach (it detects mislead-
ing assertions or claims in news information by using either deep network
models such as CNN to detect Deep syntax [64] or rhetorical structure the-
ory to detect Rhetorical structure [65] in order to check the news veracity)
or (ii). objectivity-based approach (it identifies style signs like hyper-
partisan styles and yellow journalism such as click-bait that may suggest a
deterioration in the objectivity of news material in order to mislead users).
b). Social Context Based Model: The existing approaches for constructing
social context based model of fake news detection may be either stance based
approach or propagation based approach.
•Stance based approach: This approach identifies whether the user supports,
opposes, or is neutral about a specific entity, event, or idea [67] in order
to verify the truthfulness of the news by using either direct responses like
”thumbs up” or ”thumbs down” on Facebook or automatically retrieving
from social media posts).

A Comparative Study of Computational Fake News Detection on Social Media 13
Table 2: Examples of the Fact-Checking Approaches
Example Approach Used Topics Discussed Analyzed Content
Politifact1Expert-Oriented US Politics Statements
Snopes2Expert-Oriented Politics, Social Issues News Article, Videos
Fiskkit3Crowd-sourced Politics, Society, Religion etc. Discussion Forum
FAKTA [66] Automatic MBFC website’s News media Web News based contents
1https://www.politifact.com
2https://www.snopes.com
3https://fiskkit.com
•Propagation based approach: This approach checks the veracity of news on
social media using the idea that the credibility of the related news is closely
associated with the credibility of significant posts of a social media platform.
Based on this premise, the propagation based fake news detection approach
uses two types of credibility networks to verify the truthfulness of claimed
statement in the given news - (i). homogeneous credibility networks
(include only one sort of entity, like an event or a post [58]) and (ii). het-
erogeneous credibility networks (include different sort of entities, like
events, sub-events or posts [68][69]).
2.4.3 Fake News Detection Architecture
One of the main contribution of this paper is that we have provided a well
labelled diagram of a general architecture for fake news detection which is
given in Fig.4. We have derived this architecture after going through a series
of work done so far in the area of fake news detection as discussed in Table-4
and Table-5. The main components of this architecture are as follows:
•Main Dataset: The dataset is considered as a crucial component of fake
news detection architecture as it provides data for both training and testing
fake news detection model. As we discussed earlier that fake news detection
model generally uses machine learning or deep learning algorithms to classify
the veracity of the related news on social media, the given model need to be
trained and tested in order to build an effective fake news detection system
as a whole.
•Data Preprocessing: The data preprocessing component is all responsible to
perform all preprocessing functions needed to process training data obtained
from the main dataset. It checks null or missing values in the training data
and performs preprocessing functions including (i). tokenization (it is the
process of breaking down a string of characters into smaller bits, such as
words, keywords, phrases, symbols, and other elements and is the initial step
in Natural Language Processing (NLP) [70]) and (ii). stemming (it is an
NLP method that reduces a word to its base word or stem in such a way
that comparable words are grouped together under a single stem [71]).

14 A Comparative Study of Computational Fake News Detection on Social Media
Fig. 4: A General Fake News Detection Architecture.
•Feature Extraction: This components aims to create a formal mathematical
structure in order to represent news content and related auxiliary data that
is helpful to construct fake news detection model. The feature extraction
component performs important NLP functions including (i). bag-of-words
(it is a method of counting the number of times a word appears in a docu-
ment which is helpful to compare documents and assess their similarities for
the required applications including search, classifying documents and topic
modelling [72]), (ii). n-grams (it is an n-item contiguous sequence created
from a given text sample which is helpful to extract text corpus features [73]
for building fake news detection model), (iii). TF-IDF weighting (TF-
IDF is an acronym for Term-Frequency-Inverse Document Frequency which
is a powerful technique of determining the topic of an article based on the
words it includes that measures relevance, but not frequency [74] and it has
helped to create popular and valuable tools such as Google Search [75]),
(iv). word2Vec (it is a two-layer neural network that ”vectorizes” words

A Comparative Study of Computational Fake News Detection on Social Media 15
to analyse text [76] and discover similarities mathematically by generating
extremely precise estimations about a word’s meaning on the basis of past
appearances), and (v). POS tagging (it is an accronym for Part-of-speech
tagging which is a common Natural Language Processing procedure that
involves categorising words in a text (corpus) in relation to a certain part
of speech, based on the definition of the term and its context [77]).
•Model Construction: The extracted features are fed into a constructed model
using different ideas as discussed earlier under the section of ”creation of
model” and implemented using traditional machine learning classifiers or
deep neural networks. The model is further evaluated on the basis of metrics
including (i). precision (it measures the percentage of all discovered news
that is tagged as fake), (ii). recall (it computes the percentage of tagged
fake news that are predicted as fake news), (iii). F1 score (it measures
the overall performance of a fake news detection by combining precision and
recall), and (iv). accuracy (it determines the number of correctly predicted
fake news out of tagged fake news). The model which has higher value of
precision, recall, F1 score and accuracy is selected for finally building a fake
news detection system.
•Fake News Detection System: The finally constructed fake news detection
system can be provided an online news content (related to the main dataset)
as input query to give the output in the form of either yes (confirming that
given news is fake) or no (telling that given news is true).
2.5 Fake News Detection Research Foundation
The research directions of fake news detection can be categorised into four
categories: data, feature, model and application [5] which has been refined and
properly organised in Fig.5. The idea discussed in this figure will be used as a
foundation for comparing the existing computational methods related works
of fake news detection in section-4.2.
•Data based research: It focuses on three areas of research - (a). dataset (it
aims to develop an absolute and extensive benchmark dataset with respect
to fake news); (b). temporal (it seeks to discover distinct temporal pat-
terns from authentic news in order to provide early warnings of fake news
throughout its dissemination on social media); and (c). psychological (it
tries to investigate diverse characteristics of fake news in social psychology
on a qualitative level).
•Feature based research: As the name itself suggest that feature based research
revolves around discovering worthy features for detecting fake news using
three types of major data sources for extracting features - (a). news
content,(b). social context and (c). domain specific.
•Model based research: This kind of research focuses on building empirical
as well as potential fake news detection model by applying the extracted
features into machine learning or deep learning models such as (a). super-
vised (it uses pre-annotated dataset of fake news ground truth in order to

16 A Comparative Study of Computational Fake News Detection on Social Media
Fig. 5: Research Foundation for Fake News Detection on Social Media.
train a model); (b). unsupervised (it is applicable over unlabeled fake
news dataset); (c). semi-supervised (it is applicable where limited fake
news dataset is available); or (d). deep learning or DL(a type of machine
learning that uses many layers of processing for extracting gradually higher
level features from data); followed by selection of the best performing model
classifier. It is noteworthy that supervised based model is considered to pro-
vide more accurate results for given fine-annotated ground truth dataset for
model training whereas unsupervised based model is supposed to be more
practical due to the easy availability of unlabeled dataset.
•Application based research: This research focuses on the key areas where the
fake news detection finds application like (a). fact-checking,(b). fake
news diffusion (it describes the paths of diffusion and patterns of fake
news propagation over social media sites that requires additional research,
including identifying spreaders, social factors, the life cycle and many other
things), (c). fake news intervention (it employs proactive intervention
strategies in an effort to lessen the effects of false information, as well as to
determine the best ways to intervene post the viral of fake news on social
media), (d). rumor detection,(e). stance detection (identifying the
writer of a news article is for or against a specific target) etc.

A Comparative Study of Computational Fake News Detection on Social Media 17
Fig. 6: Adopted Research Methodology for the Current Systematic Review.
3 Research Methodology
We referred the research protocol defined under PRISMA (Preferred Report-
ing Items for Systematic Reviews and Meta-analysis) [78][35] in order to
conduct the current systematic review on computational methods related
research of fake news detection on social media. The entire systematic review
is carried out in five steps which has been well described in Fig.6.:
Step-1: Organizing Published Documents According to IEC (Inclu-
sion/Exclusion Criteria)
Criteria-1: all the referred published documents are in English language;
Criteria-2: all the referred documents are original work as far as technical
aspect is concerned;
Criteria-3: all the published works emphasize on computational methods
related research on fake news detection on social media; and
Criteria-4: all the published works belong to a specific time period i.e. between
the year 2011 and April 2022;
Step-2: Selecting Resource for Availing Published Documents
•90% of the referred published documents are searched and downloaded from
Google Scholar scientific database that contains the published works from
various conferences, journals, book chapters, etc.; and
•10% of the referred published documents are downloaded from subscribed
scientific databases including Scopus, IEEE and Springer.
Step-3: Selection of Documents for Literature Review
•The referred published documents were searched with title under dou-
ble quotes ”Fake News Detection on Social Media” on scientific databases
including Google Scholar, the Scopus eSS, IEEE and Springer; and

18 A Comparative Study of Computational Fake News Detection on Social Media
•The published works with good contribution to the study of computational
methods related fake news detection on social media were selected for the
literature review.
Step-4: Collection of Data
1. First of all, we searched for the percentage of published documents under
the quoted text ”Fake News Detection on Social Media” on the Scopus eSS
scientific database1. We obtained three significant graphs as provided in
Fig.7, Fig.8 and Fig.9. These graphs help us to understand the research pub-
lication trends in the area of fake news detection on social media between
the year 2014 and the year 2021.
2. Secondly, since the research basis for reviewing the published works of com-
putational methods related to fake news detection on social media is vast,
we created three tables for data collection with different set of criteria for
comparison among the published documents.
(a) Table-3 provides comparison among the available datasets for fake news
detection;
(b) Table-4 provides a detailed comparative analysis of computational
methods related research of fake news detection on social media; and
(c) Table-5 provides comparison among the most promising empirical work
of identifying fake news on social media.
4 Detecting Fake News on Social Media using
Computational Methods: A Comparative
Study
4.1 Dataset
The datasets that have been used in existing works so far for building fake
news detection system are described in Table-3. There are three types of public
fake-news datasets: claims (they can be a single sentence or a few phrases that
include facts that should be verified, e.g. POLITIFACT, CHANNEL4.COM
and SNOPES.COM), complete articles (they are made up of a number of
sentences that are linked together to make a single piece of information, e.g.
FAKENEWSNET and BS DETECTOR), and Social Networking Services
(SNS) data (although similar in length to claims, it includes structured data
from accounts and posts, as well as a lot of non-text data, e.g. BUZZFEED-
NEWS, BUZZFACE, SOME-LIKE-IT-HOAX, PHEME and CREDBANK)
[79].
4.2 Detailed Comparison Among Existing Work
Following a thorough analysis of the gathered papers, we have tabulated
the detailed comparison among the existing computational methods based
1https://www.scopus.com/search/form.uribasic. Last retrieved on March 3, 2022

A Comparative Study of Computational Fake News Detection on Social Media 19
Table 3: Comparison among Available Fake News Detection Datasets
Dataset Volume Content Type Platform Application Mentioned in
PolitiFact 488 Claims Multiple Fake News Detection 1
LIAR 10000 (approx.) Claims Multiple Fake News Detection 2
FakeNewsNet 1000 (approx.) Complete articles Twitter Fake News Detection [5][50] [80][81] [82]
BS Detector 10000 (approx.) Complete articles Multiple Fake News Detection [5]
Twitter 1000 (approx.) SNS data Twitter Spot Fake News, Rumor [83][84]
BuzzFeedNews 2282 SNS data Facebook Fake News Detection 3
Weibo 4700 SNS data Sina Weibo Spot Fake News, Rumor [84][85][49][86][46][87]
KaggleFN 10000 (approx.) SNS data Facebook Fake News Detection [84][88][89]
PHEME 6500 SNS data Multiple Spot Fake News, Rumor [84][79][90][91][92][87]
Twitter15 1500 SNS data Twitter Spot Fake News, Rumor [46][93]
CREDBANK 60 M SNS data Twitter Credibility Assessment [94][90]
Twitter16 850 SNS data Twitter Spot Fake News, Rumor [84][46][95][93]
1[35][21][84][50][96][80][97][98] [99][100][101][81][102][50]
2[5][35][21][84][79][103][104][91][105][106][107][82][87][108]
3[5][21][35][84][79][109][90][50][96][85][91][80][97][105][81][102][82][110]

20 A Comparative Study of Computational Fake News Detection on Social Media
work done in the area of fake news detection during the time period
between the year 2011 and April 2022 in Table-4. We referred Fig.5 to use
certain criteria for comparison among the existing work in terms of refer-
enced work along with publishing year, publisher along with kind of work
(Technical Report/Book Chapter/Workshop/Conference/Journal), method of
carrying research (Survey/Experimental), kind of data-based research car-
ried (Dataset/Temporal/Psychological), kind of feature-based research carried
(Content/Context/Domain Specific), model used to train and test the dataset
(Supervised/Semi-Supervised/Unsupervised/Deep Learning) and application
of research besides fake news detection (fact-check/fake news diffusion/fake
news intervention/rumor detection/Stance detection/Others).
4.3 Comparison Among the Most Promising Empirical
Work
Table-5 provides comparative analysis of the most promising empirical works
done between the year 2011 and April 2022 in the area of fake news detec-
tion. The basis for comparison includes referenced work, input data considered
to test the model, methodology applied and the results obtained during the
experiment. This provides a good understanding about the existing empirical
works that can be considered to carry out further research in the area of fake
news detection on social media.
4.4 Discussion
Fake news detection research on social media is one of the most envisaged
research in the resent times. Post US Presidential election of 2016, the research
related work has mushroomed drastically in the area of fake news detection on
social media. This is evident from Fig.7, Fig.8 and Fig.9 that were taken from
the Scopus eSS document search1while searching for the quoted text ”Fake
News Detection on Social Media”. Fig.7 explains the number of documents
published per year between the year 2014 and the year 2021. Further, Fig.8
provides a pie-chart explaining the document types that have been published
so far. From the given pie-chart, it is evident that the Conference papers
highly contribute about 64.2% of the total published documents while Journal
Articles comes next that contribute about 27.6%, reviews contribute about
6.2% including Conference Reviews (3.7%) and review papers (2.5%). The
book chapters contribute only 1.5% of the total published work. Amazingly,
the highest number of empirical/experimental works in the area of fake news
detection have been obtained in the form of Conference paper whereas the
journal articles comes next and the book chapter contributes the least.
The leading areas in which maximum documents published related to ”Fake
News Detection on Social Media” so far can be sequenced in the decreas-
ing order as: Computer Science (42.8%), Engineering (16.4%), Mathematics
1https://www.scopus.com/search/form.uribasic. Last retrieved on March 3, 2022

A Comparative Study of Computational Fake News Detection on Social Media 21
(10.9%), Decision Science (9.8%), Social Science (5.5%), Physics and Astron-
omy (3.2%), Materials Science (2.1 %), Medicine (2.0%), Business Management
(1.9%), Energy (1.4%) and Other (3.9%). This is evident from Fig.9 which
simply explains that the social media fake news research is not just restricted
to computer science and engineering branch but it is a multidisciplinary topic
of research.
As far as the computational research work of fake news detection on social
media is concerned, a benchmark dataset is quite essential. Table-3 provides a
comparative analysis of the existing datatsets that were frequently used in the
papers analysed during the current review study. Vlachos et al. [62] are the first
to make a dataset for detecting and verifying fake news available to the public.
However, this dataset only has 221 statements obtained from CHANNEL 4 and
POLITIFACT.COM, preventing machine learning-based assessments. Wang et
al. [64] developed LIAR dataset to overcome this limitation. Further, Shu et
al. [99] developed FakeNewsNet dataset which has contributed many research
in the recent times as far as the data mining perspective is concerned with
respect to fake news detection on social media. Other significant datasets that
contribute to fake news detection research include CREDBANK [94] (a large
scale crowdsourced dataset used for accuracy assessment for Twitter events),
PHEME [111][112] (a dataset of probable rumors in Twitter and journalistic
evaluations of their veracity), Weibo [113] and Twitter [83]. In recent times,
Twitter15 and Twitter16 [114] are used in many fake news detection related
research. The limitation of existing fake news datasets is that they are not
real-time and still a lots of work has to be done in this area.
Fig. 7: Research Articles Published Yearly related to Fake News Detection on
Social Media.
Table-4 provides comparative Analysis of existing Work in Computa-
tional Methods of Fake News Detection on Social Media. Around 63 papers

22 A Comparative Study of Computational Fake News Detection on Social Media
Fig. 8: Types of Documents Published related to Fake News Detection on
Social Media.
Fig. 9: Subject Areas related to Fake News Detection on Social Media in
which Documents Published.
were critically reviewed to provide a detailed analysis from the time com-
putational methods for fake news detection related research firstly done in
the year 2011 by Castillo et. al. [12] to the recent work done by Min et.
al. [115]. The comparison is basically done on the basis of reference along
with the published year, publisher along with the kind of work carried
(Technical Report/Book Chapter/Workshop/Conference/Journal), method of
carrying research (Survey/Experimental), kind of data-based research car-
ried (Dataset/Temporal/Psychological), kind of feature-based research carried
(Content/Context/Domain Specific), model used to train and test the dataset

A Comparative Study of Computational Fake News Detection on Social Media 23
(Supervised/Semi-Supervised/Unsupervised/Deep Learning) and application
of research besides fake news detection (fact-check/fake news diffusion/fake
news intervention/rumor detection/Stance detection/Others).
Table-5 provides comparative anlaysis of the most promising empirical
works done in the area of fake news detection on social media. Around 41 most
promising empirical works done between the year 2011 and April 2022 have
been included in this table. The basis for comparing these empirical works
include the kind of input data taken during training and testing fake news
detection model, the methodology used to carry out the empirical research and
the results obtained during the experiment of the given fake news detection
system.
5 Major Research Findings
Early efforts to automatically detect fake news were primarily focused on cre-
ating helpful features from a range of information sources, including textual
content [12] [140][141], publisher’s personal details [12], [17] and commu-
nication means [142][143][144]. However, the fact that these feature-based
algorithms take more time and involve more work poses a serious problem.
Further, they are not considered ideal in most of the cases due to high depen-
dency of the model on the quality of artificial features. As a result of the
model’s heavy dependence on the quality of artificial features, they are also
not often regarded as ideal one.
The main issues that prevent the effectiveness of the existing fake news
detection methods are related to the extremely adaptable character of false
information. A generic dataset for fake news identification is, in fact, quite
difficult to obtain. As a result, it is highly challenging to extract pertinent
features that accurately describe and enable the detection of fake news across
myriad domains. However, the efforts have been made in the recent times to
come out with benchmark datasets including PolitiFact1, Twitter [83] Buz-
zFeedNews [145], LIAR [64], Weibo [113], FakeNewsNet [99], Twitter15 and
Twitter16 [114] to carry out fake news detection related research. However,
these datasets cover the fake news related details belonging to a particular
region, state or country. There is scarcity of benchmark datasets that cover
the global incidents related to fake news.
In various works employing deep neural networks, a variety of neural net-
work related models have been used. For instance, Ma et. al [13] employed
recurrent neural network (RNN) to determine the best way to display tweet
text on the posting timeline. Ruchansky et. al [57] developed a model called
CSI (which stands for Capture, Score, Integrate) that combines all three fea-
tures (specifically an article’s text, the response received by user, and the users
who are promoting it from the source) in order to make a more accurate and
automated prediction of fake news. Zhang et. al [126] constructed a deep diffu-
sive network model called FakeDetector to simultaneously discover how news
1https://www.politifact.com

24 A Comparative Study of Computational Fake News Detection on Social Media
Table 4: Computational Methods Related Research of Fake News Detection on Social Media
Reference Publisher Method Data1Feature2Model3Application4
Castillo et al. [12] (2011) ACM (Conference) Experimental (b) (a),(b),(c) (a) (a)
Vlachos et al. [62] (2014) ACL (Workshop) Survey (a) (a),(b) — (b)
Rubin et al. [116] (2015) IEEE (Conference) Experimental (a),(c) (b) (b) (a)
Pennebaker et al. [117] (2015) Univ. of Texas Survey (c) (a),(b) — —
Ferreira et al. [118] (2016) ACL(Conference) Experimental (a) (a),(b) — (b)
Jin et al. [58] (2016) AAAI (Conference) Experimental (b) (c) (b) (a)
Ozgobek et al. [119] (2017) NOBIDS (Conference) Survey — (a),(b),(c) — (a)
Shu et al. [5] (2017) ACM (Journal) Survey (a),(b),(c) (a),(b),(c) (a),(b),(c) (a),(d)
Janze et al. [109] (2017) AIS (Conference) Experimental (a) (a) (a),(d) (a)
Wang et al. [64] (2017) arXiv (Journal) Experimental (a) (a),(b),(c) (a) (a)
Buntain et al. [90] (2017) IEEE (Conference) Experimental (a),(b),(c) (a),(b),(c) (a), Other (c)
Shu et al. [120] (2017) arXiv (Journal) Experimental (a) (a),(b) (c) (a)
Ruchansky et al. [57] (2017) ACM (Conference) Experimental (b) (a),(b),(c) (d) (a)
Oshikawa et al. [79] (2018) arXiv (Journal) Experimental (a) (a) (a),(b),(c) (a)
Kotteti et al. [103] (2018) IEEE (Conference) Experimental (a) (a) (a) —
Aldwairi et al. [11] (2018) Elsevier (Conference) Experimental (b) (a) (a) (e)
Pan et al. [88] (2018) Springer (Conference) Experimental (a) (a) (a) (a)
Hanselowski et al. [121] (2018) arXiv (Conference) Experimental (a) (a) (a),(d) (d)
1Data based research:- (a). Dataset, (b). Temporal, and (c). Psychological
2Feature based research:- (a). News Content, (b). Social Context, and (c). Domain Specific
3Model based research:- :- (a). Supervised, (b). Unsupervised, (c). Semi-supervised, and (d). Deep Learning
3Application based research:- (a). Fake News Diffusion, (b). Fact Check, (c). Rumor Detection, (d). Stance Detection, and (e).
Fake News Intervention

A Comparative Study of Computational Fake News Detection on Social Media 25
Table 4: (Continued)
Reference Publisher Method Data1Feature2Model3Application4
Della et al. [96] (2018) IEEE (Conference) Experimental (a) (a),(b) (a) (a)
Atodiresei et al. [122] (2018) Elsevier (Conference) Experimental (b),(c) (a),(b) Other (a)
Fernandez et al. [104] (2018) Springer (Conference) Experimental (a) (a),(b) (d) (a)
Parikh et al. [91] (2018) IEEE (Conference) Experimental (a),(c) (a),(b) (a),(b) (a)
Shu et al. [99] (2018) arXiv (Journal) Experimental (a),(b) (a),(b),(c) (a),(d) (a),(e)
Wang et al. [49] (2018) ACM (Conference) Experimental (b) (a) (d) (a),(e)
Shu et al. [80] (2018) IEEE (Conference) Experimental (a),(c) (a),(b),(c) (a),Other (a)
Ajao et al. [92] (2018) ACM (Conference) Experimental (a),(b) (a),(b),(c) (d),Other —
Meinertet et al. [123] (2018) Springer (Conference) Survey (a),(c) (a),(b),(c) (a),Other (a),(b)
Kumar et al. [124] (2018) arXiv (Journal) Survey (a),(b),(c) (a),(b),(c) (a),(b),(c) (a),(b)
Karimi et al. [106] (2018) ACL (Conference) Experimental (b) (a),(b) (a),(d),Other —
De et al. [125] (2018) arXiv (Journal) Experimental (a),(b) (a),(b) (c) (e)
Gupta et al. [97] (2018) IEEE (Conference) Experimental (c) (a),(b) (a),(d),Other (a)
Dong et al. [105] (2018) Springer (Conference) Experimental (b) (a),(b) (d) —
Zhang et al. [126] (2018) arXiv (Journal) Experimental — (a),(b) (d) (a)
Wu et al. [127] (2018) ACM (Conference) Experimental (b) (a) (a) (a)
Guacho et al. [128] (2018) IEEE (Conference) Experimental (a) (a) (c) —
Helmstetter et al. [129] (2018) IEEE (Conference) Experimental — (a) (a) —
1Data based research:- (a). Dataset, (b). Temporal, and (c). Psychological
2Feature based research:- (a). News Content, (b). Social Context, and (c). Domain Specific
3Model based research:- :- (a). Supervised, (b). Unsupervised, (c). Semi-supervised, and (d). Deep Learning
3Application based research:- (a). Fake News Diffusion, (b). Fact Check, (c). Rumor Detection, (d). Stance Detection, and (e).
Fake News Intervention

26 A Comparative Study of Computational Fake News Detection on Social Media
Table 4: (Continued)
Reference Publisher Method Data1Feature2Model3Application4
Liu et al. [46] (2018) AAAI (Conference) Experimental (b) (a) (d) —
Sharma et al. [84] (2019) ACM (Journal) Survey (a),(b),(c) (a),(b),(c) — (a),(e)
Shu et al. [100] (2019) ACM (Conference) Experimental — (a),(b) (a) —
Olivieri et al. [107] (2019) AIS (Conference) Experimental — (a),(b) (a) —
Pierri et al. [21] (2019) ACM (Journal) Survey (a) (a),(b) — (a),(e)
Reis et al. [110] (2019) IEEE (Journal) Experimental (b) (a),(b),(c) (a) —
Guo et al. [130] (2019) arXiv (Journal) Experimental — (a),(b) (d) (a)
Bharadwaj et al. [89] (2019) HAL (Journal) Experimental — (a) (a),(d) —
Cardoso et al. [87] (2019) AIS (Conference) Survey (a) (a),(b) (a),(b),(d) —
Rasool et al. [108] (2019) ACM (Conference) Experimental — (a) (a) (a),(e)
Traylor et al. [131] (2019) IEEE (Conference) Experimental — (a) (a) Spot Propaganda
Shu et al. [50] (2019) ACM (Conference) Experimental (b),(c) (a),(b) (c) —
Shu et al. [81] (2019) Springer (Journal) Experimental (b) (a),(b) (a),(e) (a)
Yang et al. [82] (2019) AAAI (Conference) Experimental (b) (a) (b) (a)
Shu et al. [101] (2020) AAAI (Conference) Experimental (b) (c) (b) (a)
Zhang et al. [45] (2020) Elsevier (Journal) Survey (a),(b),(c) (a),(b),(c) — (a),(e)
Zhou et al. [132] (2020) ACM (Journal) Experimental (c) (a),(b),(c) (a),(d) (a)
Lu et al. [95] (2020) ACL (Conference) Experimental — (a) (d) —
1Data based research:- (a). Dataset, (b). Temporal, and (c). Psychological
2Feature based research:- (a). News Content, (b). Social Context, and (c). Domain Specific
3Model based research:- :- (a). Supervised, (b). Unsupervised, (c). Semi-supervised, and (d). Deep Learning
3Application based research:- (a). Fake News Diffusion, (b). Fact Check, (c). Rumor Detection, (d). Stance Detection, and (e).
Fake News Intervention

A Comparative Study of Computational Fake News Detection on Social Media 27
Table 4: (Continued)
Reference Publisher Method Data1Feature2Model3Application4
Kesarwani et al. [133] (2020) IEEE (Conference) Experimental — (a) (a) —
Gimpel et al. [134] (2020) HICSS (Conference) Experimental — (a),(b) Other —
Ni et al. [93] (2021) IEEE (Journal) Experimental (b) (a),(c) (d) (a),(e)
Wang et al. [135] (2021) arXiv (Journal) Experimental — (b) (d) —
Sahoo et al. [136] (2021) Elsevier (Journal) Experimental (b) (a) (d) (a)
Jarrahi et al. [137] (2021) ACM (Conference) Experimental — (c) (d) (a)
Shu et al. [138] (2022) Springer (Book Chapter) Experimental (b) (a),(c) (d) (a)
Seddari et al. [139] (2022) IEEE (Journal) Experimental (c) (a),(b),(c) (a) (a)
Min et al. [115] (2022) ACM (Conference) Experimental (a),(b) (a) (d) (a)
1Data based research:- (a). Dataset, (b). Temporal, and (c). Psychological
2Feature based research:- (a). News Content, (b). Social Context, and (c). Domain Specific
3Model based research:- :- (a). Supervised, (b). Unsupervised, (c). Semi-supervised, and (d). Deep Learning
3Application based research:- (a). Fake News Diffusion, (b). Fact Check, (c). Rumor Detection, (d). Stance Detection, and (e).
Fake News Intervention

28 A Comparative Study of Computational Fake News Detection on Social Media
sources, authors, and topics are represented on the basis of extraction of a
collection of latent and explicit features from the text based information. Liu
et. al [46] employed RNN (Recurrent Neural Network) and CNN (Convolution
Neural Network) in tandem to track changes in user features throughout the
propagation path, while modelling it as a multivariate time series. Guo et. al
[130] built an EFN framework for spotting fake news based on emotion after
learning how to represent the news content and its corresponding comment-
emotion with respect to publishers and consumers. Lu et. al [95] developed a
GCAN (Graph-aware CoAttention Network) to verify the veracity of a source
tweet using the tweet’s short content and the order in which individuals retweet
it without text comments. Ni et. al [93] created MVAN (Multi-View Attention
Networks) model to acquire significant hidden clues in the original tweet text
and the propagation structure simultaneously to detect fake news. Wang et. al
[135] proposed a BERT (Bidirectional Encoder Representations from Trans-
formers) model using BiLSTM layers and CNN layers on top for COVID-19
related fake news detection. Min et. al [115] built a PSIN (Post-User Inter-
action Network) using a divide-and-conquer approach to effectively represent
post-user, user-user and post-post interactions in terms of the social context
while working over their intrinsic properties for spotting fake news. Most of
these approaches, however, have a major drawback that they only work with
sequence data but not with structured data as a result of which they are very
less effective to simulate the propagation structure in real life.
Among the publications from Table-4, the key issue with the majority of
them is that while they work well when used with a particular input dataset,
they are not effective enough to get generalised to missing data. A very few
works are there including Kotteti et. al [103], Wu et. al [127] and Olivieri et. al
[107] that address this issue. However, still more reliable and effective works are
required in this area. According to our analysis, methodologies that are only
focused on content analysis appear to have a limited application, context anal-
ysis targets more easily generalizable acts (including liking, commenting, and
spreading) while domain specific appear to be more effective to detect newly
emerged news content on social media. However, we analysed a very limited
publications employing domain specific methodology for fake news detection
on social media. On the other hand, 90% of the works have utilised either
content based or social context based or both methodologies. Moreover, most
of the documents analysed in Table-4 finds application in fake news diffusion
while a very few empirical based publications contribute to the area of (i).
fake news intervention including that of Aldwairi et. al [11], Wang et. al
[49], De et. al [125], Ni et. al [93]; (ii). fact-checking by vlachos et. al [62],
Ferreira et. al [118]; (iii). rumor detection by Buntain et. al [90]; and (iv).
stance detection by Hanselowski [121].
Among the publications from Table-5, we found the works of Liu et al. [46],
Aldwairi et al. [11], Pan et al. [88], Wang et al. [49], Wu et al. [127], Lu et
al. [95] and Sahoo et al. [136] as the most promising as far as content based
model is concerned; the works of Rubin et al. [116] and Wang et al. [135] as

A Comparative Study of Computational Fake News Detection on Social Media 29
the most promising when context based model is considered; and the most
promising domain based model is developed by Jarrah et al. [137]. There
are a large number of existing empirical based models that combine content,
context and domain to provide an effective fake news detection mechanism on
social media including Castillo et al. [12], Wang et al. [64], Buntain et al. [90],
Ruchansky et al. [57], Ajao et al. [92], Zhou et al. [132] and Seddari et al. [139].
One of the leading challenges in the existing computational method based
fake news detection research is that a lots of works are still to be done in
order to spot newly emerged fake news. It is very difficult to identify fake
news related to the recent occurred news event due to the lack of knowledge
and information associated with it. Because of this, using a single technique
to spot fake news articles won’t be as effective as needed. There are very few
promising empirical based publications to address this issue including Aldawiri
et al. [11], Wang et al. [49] and de Alfaro et al. [125]. Further, the performance
of fake news detection models can be enhanced by the attention mechanism. A
very limited work has been done so far in this area where the most promising
works done so far includes Dong et al. [105], Lu et al. [95] and Ni et al. [93].
6 Research Gaps and Future Directions
After reviewing around 63 papers whose comparative analysis are provided in
Table-4 and Table-5 respectively on the basis of different set of criteria, we
found following significant research gaps that could become future scope of
research in the area of computational methods based fake news detection on
social media related research:
•One of the major limitations of existing research is that there is still a lots
of work required to develop method for accurately detecting fake news at
early stage before it goes viral on social media so that necessary steps can
be taken to mitigate and intervene the fake news spread. In short, there is
utmost need of real time fake news detection system. Zhou and Zafarani
[22] proposed three ways to do so: (i). dynamic Knowledge Graph related
methods shall be built to help the timely updates of the real-word data
obtained from the newly emerged news on social media which is otherwise
difficult to get stored in existing knowledge graphs [22]; (ii). selection of
those features shall be done for building fake news detection model that
can identify how misleading writing has changed through time and how it is
common across themes, languages, domains, and likes [49]; and (iii). either a
less information such as headlines shall be used or topics of worth-checking
shall be picked for efficient news verification [46][132].
•While the research using a conventional machine learning framework with
theory-based or pattern-based feature engineering to identify fake news have
already been conducted, there hasn’t been much research carried using
domain specific ideas to direct the machine learning process or deep neural
networks to detect fake news.

30 A Comparative Study of Computational Fake News Detection on Social Media
Table 5: Comparison among the Most Promising Empirical Work of Fake News Detection on Social Media
Reference Input Data Methodology Results
Castillo et al.
[12] (2011)
Twitter Event Developed a decision tree-based model to pre-
dict news credibility using a large feature set
The supervised classifier achieves
an accuracy of 86%
Rubin et al.
[116] (2015)
News transcripts with
associated RST analyses
Used SVM classifier along with mixed
approaches of RST and VSM to classify a news
as real or fake
The supervised classifier achieves
an accuracy of 86%
Ferreira et
al. [118]
(2016)
Rumoured claims, news
articles from Emergent
dataset
Developed a model for stance classification
using multiclass logistic regression
The accuracy achieved was 73%
which was 26% better than one
achieved by EOP
Jin et al. [58]
(2016)
Tweets from Sina Weibo Used conflicting social viewpoints in a credi-
bility propagation network for verifying news
automatically in microblogs
The proposed model called
CPCV achieved accuracy of 84%
Janze et al.
[109] (2017)
Fact-checked data of
Facebook Post from
BuzzFeed dataset
Developed a model for exploratory research,
based on mixed approaches of ELM and UGC,
using visual, behavioral, cognitive and emo-
tional cues of the news posted on Facebook
The best performing configura-
tions achieve a predictive accu-
racy of more than 80%
Wang et al.
[64] (2017)
Short statements from
LIAR dataset
Used 5 baselines (majority baseline, regular-
ized LR, SVM, Bi-LSTMs, CNNs) on surface-
level linguistic patterns in LIAR
CNNs outperformed all models,
resulting in an accuracy of 0.270
on the heldout test set
Buntain
et al. [90]
(2017)
Tweet Events from
CREDBANK and
PHEME datasets
Created a classification model to predict a
Twitter thread post to be true or false
The proposed models perform
better than [146] with 61.81%
accuracy as compared to 66.93%
and 70.28% accuracy in PHEME
and CREDBANK respectively
Shu et al.
[120] (2017)
News articles from Fake-
NewsNet dataset
Developed TriFN framework to spot fake news
by simultaneous modelling of publisher-news
relationship and user-news discourse
TriFN outperforms LIWC [117]
plus Castillo [12] by accuracy
of 4.72% (BuzzFeed) and 5.84%
(PolitiFact) respectively

A Comparative Study of Computational Fake News Detection on Social Media 31
Table 5: (Continued)
Reference Input Data Methodology Results
Ruchansky et
al. [57] (2017)
Tweet Events from
Twitter and Weibo
datasets
Created CSI model to capture three specific
traits of fake news (text, response, and source)
in order to spot false information w.r.t the user
and article
The CSI model achieved accu-
racy and F1 score of 89.2% and
89.4% respectively (on Twitter
dataset) while 95.3% and 95.4%
respectively (on Weibo dataset)
Kotteti et al.
[103] (2018)
Short statements from
LIAR dataset
Used data imputation for both categorical and
numerical features in order to solve the missing
values issue in a dataset; besides using TF-IDF
vectorization to remove unimportant features
The supervised classifier achieves
an accuracy of 86%
Aldwairi et al.
[11] (2018)
Social Media Web
pages
Proposed a basic tool installation into personal
browser in order to detect and filter out poten-
tial clickbaits
Achieves a 99.4% accuracy using
the logistic classifier
Pan et al. [88]
(2018)
News articles Developed a B-TransE model using the mixed
approaches of knowledge graphs and the
TransE model
Achieves F1 ratings of over 0.80
in some of the outputs
Della et al.
[96] (2018)
Facebook posts and
their likes
Developed a HC-CB-3 to automatically spot
fake news and tested it within chatbots using
vivid datasets
Achieved accuracy of 81.7%
Wang et al.
[49] (2018)
Tweet from Twitter
and Weibo datatsets
Designed an event-invariant framework
called the Event Adversarial Neural Network
(EANN), which includes a multimodal feature
extractor, a fake news detector, and an event
discriminator for fake news detection of newly
arriving events
Achieved accuracy of 71.5% and
82.7% on Twitter and Weibo
dataset respectively
Ajao et al. [92]
(2018)
Tweets from works by
[147]
Developed a hybrid deep learning model using
LSTM and CNN models to automatically spot
features inside posts
Achieved 82% accuracy without
any prior knowledge of the sub-
jects being addressed

32 A Comparative Study of Computational Fake News Detection on Social Media
Table 5: (Continued)
Reference Input Data Methodology Results
Karimi et al.
[106] (2018)
Short statements
from LIAR dataset
Designed a MMFD framework using CNN and
LSTM using mixed approaches of automated
feature retrieval, multi-source based fusion
and automated fake detection
Achieves the highest accuracy of
38.81% out of different sources combi-
nations
De et al.
[125] (2018)
News posted on
Twitter
Developed reputation systems for news shared
on Twitter using three distinct methods
and employing logistic regression classifier for
training and testing dataset
Detected fake URLs at the rate of over
80% on complete training sets while
70% for the Min-2 training set
Wu et al.
[127] (2018)
Tweets from Twit-
ter
Developed a TraceMiner model to spot fake
news using diffusion network details even if
content details are missing
TraceMiner outperformed all baselines
like TM(DeepWalk), TM(LINE) for
sample ratios - 0.1 to 0.9
Gupta et al.
[97] (2018)
News articles Modeled the echo-chambers within the social
network to represent a news piece as a 3-mode
tensor with the structure ”News, User, Com-
munity” while latently embedding the news
article
The CITDetect model achieves ideal
recall and high F1-score of 1.000 ±
0.000 and 0.813 ±0.078 respectively
on Buzzfeed dataset
Dong et al.
[105] (2018)
Short statements
from LIAR, Buz-
zFeedNews
Developed a DUAL model using a bi-
directional attention-based GRU for feature
extraction that is later used to create an atten-
tion matrix in order to spot fake news
Achieved accuracy of 82.3% on LIAR
dataset and 83.8% on BuzzFeedNews
dataset
Zhang et al.
[126] (2018)
Tweets and fact-
check articles from
PolitiFact
Created a FakeDetector model to determine
fake news credibility using a deep diffusive net-
work model
Achieved 14.5% higher accuracy than
fake news detection models (Hybrid
CNN, LIWC, TRIFN), network
structure-based models (PROPA-
GATION, DEEPWALK, LINE) and
text-based methods (RNN, SVM)

A Comparative Study of Computational Fake News Detection on Social Media 33
Table 5: (Continued)
Reference Input Data Methodology Results
Guacho et al.
[128] (2018)
News articles from
public dataset
Built a KNN graph of news pieces using embed-
dings of tensor-based articles in order to spot
their similarity in an embedding, latent space
The proposed technique outper-
forms SVM classifier by 75.43%
despite only employing 30% of the
labels from a public dataset
Liu et al. [46]
(2018)
Sharing cascades on
Weibo, Twitter15 and
Twitter16 datasets
Built a time series classifier using RNN and
CNN in order to spot fake news by retrieving
the global and local changes in user’s identity
along the propagation path
Outperformed all baselines to spot
fake news at the early stage at
accuracy of 85% and 92% on Twit-
ter and Sina Weibo respectively
Shu et al. [100]
(2019)
Claims from Politi-
fact and Gossipcop
datasets
Measure users’ sharing behaviours to identify
group representative with high chance to share
fake news; and comparative analysis of their
explicit and implicit profile features, in order to
spot fake news
Achieved higher accuracy of 90.0%
(PolitiFact), 96.6% (Gossipcop)
while considering all profile fea-
tures as compared to when only
explicit/implicit features included
Reis et al.
[110] (2019)
News articles Supervised learning classifiers are applied over
dataset with proposed key features
Random forests, XGBoost outper-
form all baselines in terms of ROC
Curve and F1-score
Olivieri et al.
[107] (2019)
News Articles from
PolitiFact
Task-generic features (extracted using Google’s
search engine answers while crowdsourcing for
missing details) are combined with textual fea-
tures to spot fake news
Results demonstrate a consider-
able increase in F1-Score for a
6-class task of 3% over the state of
the art methods
Guo et al.
[130] (2019)
Sharing cascades from
Weibo dataset
Developed a deep learning based EFN frame-
work that uses both content and social emotions
simultaneously to spot fake news
EFN achieved accuracy of 87.2%,
12% higher than decision trees
plus 4% higher f1-score
Bharadwaj et
al. [89] (2019)
News articles Used semantic features along with several
machine learning methods to detect online fake
news
An accuracy of 95.66% was
attained using bigrams and ran-
dom forest

34 A Comparative Study of Computational Fake News Detection on Social Media
Table 5: (Continued)
Reference Input Data Methodology Results
Rasool et al.
[108] (2019)
Short statements
from LIAR dataset
Multilevel multiclass method for fake news
detection using dataset relabeling and iterative
learning
The proposed technique performs
better than the benchmark with
an accuracy of 39.5%
Traylor et al.
[131] (2019)
News Articles Developed a system using Textblob, NLP,
SciPy Toolkits through a quoted attribution
process (called influence mining) in a Bayesian
machine learning algorithm to spot fake news
The precision of the resulting pro-
cess is 63.333% in determining if a
quote-heavy article is likely to be
fake
Shu et al. [81]
(2019)
Short statements Built a FakeNewsTracker model using deep
LSTMs with two layers, hundred cells at each
layers of both encoder and decoder
Achieved accuracy of 67% on Poli-
tiFact) and 74.2% on BuzzFeed
Yang et al. [82]
(2019)
Short statements
from LIAR and Buz-
zFeedNews datasets
Built a probabilistic graphical model called
UFD using extracted social media users’ opin-
ions from the data of their social engagement
hierarchy and Gibbs sampling approach to spot
fake news and to determine users’ credibility
Achieved accuracy of 75.9% on
LIAR dataset, 67.9% on Buz-
zFeedNews dataset
Kesarwani et
al. [133] (2020)
Short statements
from BuzzFeedNews
K-Nearest Neighbor classifier is employed to
detect fake news on social media
Achieved 79% classification accu-
racy
Shu et al. [101]Short statements Compared hierarchical propagation network
features of fake/real news from structural/tem-
poral/linguistic aspects and experimentally
checked their role in spotting fake news
The STFN-HPFN achieves high-
est accuracy of 85.6% on Poli-
tiFact dataset as compared to
86.3% on Gossipcop dataset
Zhou et al.
[132] (2020)
Short statements
from PolitiFact
and BuzzFeedNews
datasets
Presented a theory-driven model based on
supervised learning to examine news content
at the levels of lexicon, syntax, semantic, dis-
course, social and forensic psychology theories
to spot fake news
Achieved accuracy of 60% to 70%
on PolitiFact, 50% to 60% on Buz-
zFeed.

A Comparative Study of Computational Fake News Detection on Social Media 35
Table 5: (Continued)
Reference Input Data Methodology Results
Lu et al. [95]
(2020)
Sharing cascades
from Twitter15 and
Twitter16 datasets
Created a neural network-based Graph-aware
CoAttention Networks (GCAN) model to spot
fake tweet given the source tweet and the
matching retweet sequence without comments
Outperformed all the baselines
with an accuarcy of 87.67% on
Twitter15 as compared to 90.84%
accuracy on Twitter16 dataset
Ni et al. [93]
(2021)
Sharing cascades
from Twitter15 and
Twitter16 datasets
Developed a deep learning based MVAN
(Multi-View Attention Networks) model to
spot fake news on early basis by merging two
attention mechanisms - text semantic attention
and propagation structure attention so as to
acquire key hidden clues in the source tweet and
the propagation structure at the same time
Outperforms all baselines by
achieving the accuracy of 92.34%
on Twitter15 dataset as compared
to 93.65% on Twitter16 dataset
Wang et al.
[135] (2021)
Short statements
from COVID-19
dataset from Kaggle
Applied Bidirectional Encoder Representations
from Transformers (BERT) model in combina-
tion of BiLSTM, CNN layers on top with both
frozen and not frozen parameters methods to
spot COVID-19 related fake news
Achieved the highest accuracy
of 93.47% while testing various
BERT related proposed models
Jarrahi et al.
[137] (2021)
News articles from
FakeNewsNet dataset
Developed UPFD framework to spot fake news
using joint content and graph modelling to col-
lect different signals from user preferences
Achieved the accuracy of 90.6% on
PolitiFact dataset, 97.8% on Gos-
sipcop dataset
Sahoo et al.
[136] (2021)
News articles from
FakeNewsNet dataset
Developed a mechanism in the Chrome envi-
ronment to spot false news on Facebook by
analyzing the account’s activity using LSTM
The proposed deep learning algo-
rithm achieved 99.4% accuracy
Seddari et al.
[139] (2022)
Short statements
from BuzzFeedNews
datasets
Proposed a hybrid system that combines lan-
guage and knowledge-based strategies to iden-
tify fake news
The accuracy of the proposed sys-
tem using both types of features is
94.4%

36 A Comparative Study of Computational Fake News Detection on Social Media
•A fake news intervention technique shall be based either on users or on
network structure. From the viewpoint of the user, fake news intervention
depends on particular roles people play in the spread of fake news [22] - (i).
powerful spreaders of fake news on a social media if they are stopped will
result in a more effective intervention rather than focusing on individuals
who have little or no social impact; (ii). fact-checker (one who intervenes
to stop the spread of false information on social media) by including links
in their posts or comments; (iii). adversarial user (one who is against the
given opinion) shall be identified and be penalized if found guilty of spread-
ing fake news; and (iv). normal users (one who is not involved in fake news
spread) shall be be given assistance to make it easier for them to spot false
news. As comparison to the manual approach of user based fake news inter-
vention, network based fake news intervention is automatic that blocks the
propagation paths of fake news in an effort to stop it from spreading. It does
this by examining the network structure of the fake news’s propagation and
forecasting how it will spread in the future. A lots of work is supposed to
be done in this area.
•Fake news is a global phenomenon and it effects the entire world. Each
country on the globe has its own social, political and technical challenges
to address the fake news issue which is required to be explored in future.
Moreover, a benchmark dataset addressing the local/national incidents shall
be developed to further carry fake news related empirical work.
•The majority of the existing datasets categorise information as true or false
using binary labels. However, a more precise categorization of data by intent
[84] may be particularly useful in differentiating genuine fake news from
similarly related information like opinion and satire news.
•Apart from machine learning and deep learning based fake news detec-
tion techniques, the emphasise shall be given on other reliable methods of
identifying fake news more effectively and efficiently.
•Fake news detection research finds application in various areas including
rumor detection, spammer and bot detection, stance detection, clickbait
detection, truth discovery (i.e. the issue of distinguishing fact from fiction in
the presence of contradictory information) [5]. These areas demand a lots of
work to be done in future in order to address issues associated with them.
•Big fake news data and the best way to reduce feature vector size are the
hot topics that requires significant research attention [35].
•There is not any reliable work done in the area of visual based feature
extraction from visual entities such as images and videos and building its
related fake news detection system.
•The performance of models for identifying fake news can be enhanced with
the use of the attention mechanism. A lots of promising work is yet to be
done in this area.

A Comparative Study of Computational Fake News Detection on Social Media 37
7 Conclusion
This survey paper thoroughly examines and assesses the existing studies on
fake news detection on social media during the period between the year 2011
and April 2022 by (i). providing an organized and detailed theoretical back-
ground of fake news detection research on social media that covered definition
of fake news in all aspects (including its related concepts, its broad definition as
well as its narrow definition), its effects on real world, its 5Vs as five challenges
for fake news detection on social media, its detection mechanism including
mathematical modelling, detection framework, detection architecture, and its
future research foundation; (ii). doing comparative study of existing fake news
datasets, detailed comparison among 63 existing fake news detection related
research works, systematic comparison of 41 most promising empirical works
that exist during this period and discussing the key points derived from the
above various comparative studies; (iii). presenting major research findings
from the current review study; and (iv). pointing out the major research gaps
and future scope of research in the area of fake news detection on social media.
Acknowledgments. We owe a great deal of gratitude to all the authors and
researchers whose research papers and related works have given us deep insight
to write this review article. We want to further extend our gratitude to Mr.
Pankaj Singh (Alumnus, National Institute of Technology, Bhopal, India) and
Dr. Arun Aggarwal (Assistant Professor, Department of Computer Science,
Ramanujan College, New Delhi, India) for the timely and valuable research
assistance provided during the writing of this paper.
Declarations
This work is neither sponsored nor funded by any organisation. The figures
included in this review article are unique that do not require any approval. All
the works which have been referred are properly cited and given proper credits.
References
[1] Kietzmann, J.H., Hermkens, K., McCarthy, I.P., Silvestre, B.S.: Social
media? get serious! understanding the functional building blocks of social
media. Business horizons 54(3), 241–251 (2011)
[2] Allcott, H., Gentzkow, M.: Social media and fake news in the 2016
election. Journal of economic perspectives 31(2), 211–36 (2017)
[3] Perrin, A.: Social media usage. Pew research center 125, 52–68 (2015)
[4] Howell, L., et al.: Digital wildfires in a hyperconnected world. WEF
report 3(2013), 15–94 (2013)

38 A Comparative Study of Computational Fake News Detection on Social Media
[5] Shu, K., Sliva, A., Wang, S., Tang, J., Liu, H.: Fake news detection on
social media: A data mining perspective. ACM SIGKDD explorations
newsletter 19(1), 22–36 (2017)
[6] Shao, C., Ciampaglia, G.L., Varol, O., Yang, K.-C., Flammini, A.,
Menczer, F.: The spread of low-credibility content by social bots. Nature
communications 9(1), 1–9 (2018)
[7] Vosoughi, S., Roy, D., Aral, S.: The spread of true and false news online.
Science 359(6380), 1146–1151 (2018)
[8] Carson, A., Fallon, L.: Fighting fake news: a study of online misinforma-
tion regulation in the asia pacific (2021)
[9] Fisher, M., Cox, J.W., Hermann, P.: Pizzagate: From rumor, to hashtag,
to gunfire in dc. Washington Post 6, 8410–8415 (2016)
[10] Tandoc Jr, E.C., Lim, Z.W., Ling, R.: Defining “fake news” a typology
of scholarly definitions. Digital journalism 6(2), 137–153 (2018)
[11] Aldwairi, M., Alwahedi, A.: Detecting fake news in social media net-
works. Procedia Computer Science 141, 215–222 (2018)
[12] Castillo, C., Mendoza, M., Poblete, B.: Information credibility on twitter.
In: Proceedings of the 20th International Conference on World Wide
Web, pp. 675–684 (2011)
[13] Ma, J., Gao, W., Mitra, P., Kwon, S., Jansen, B.J., Wong, K.-F., Cha,
M.: Detecting rumors from microblogs with recurrent neural networks
(2016)
[14] Ozbay, F.A., Alatas, B.: Fake news detection within online social media
using supervised artificial intelligence algorithms. Physica A: Statistical
Mechanics and its Applications 540, 123174 (2020)
[15] Ahmed, H., Traore, I., Saad, S.: Detection of online fake news using
n-gram analysis and machine learning techniques. In: International Con-
ference on Intelligent, Secure, and Dependable Systems in Distributed
and Cloud Environments, pp. 127–138 (2017). Springer
[16] Riedel, B., Augenstein, I., Spithourakis, G.P., Riedel, S.: A simple but
tough-to-beat baseline for the fake news challenge stance detection task.
arXiv preprint arXiv:1707.03264 (2017)
[17] Yang, F., Liu, Y., Yu, X., Yang, M.: Automatic detection of rumor on
sina weibo. In: Proceedings of the ACM SIGKDD Workshop on Mining
Data Semantics, pp. 1–7 (2012)

A Comparative Study of Computational Fake News Detection on Social Media 39
[18] Meyers, M., Weiss, G., Spanakis, G.: Fake news detection on twitter using
propagation structures. In: Multidisciplinary International Symposium
on Disinformation in Open Online Media, pp. 138–158 (2020). Springer
[19] Freire, P.M.S., Goldschmidt, R.R.: Fake news detection on social media
via implicit crowd signals. In: Proceedings of the 25th Brazillian Sym-
posium on Multimedia and the Web, pp. 521–524 (2019)
[20] Niklewicz, K.: Weeding out fake news: an approach to social media
regulation. European View 16(2), 335–335 (2017)
[21] Pierri, F., Ceri, S.: False news on social media: a data-driven survey.
ACM Sigmod Record 48(2), 18–27 (2019)
[22] Zhou, X., Zafarani, R.: A survey of fake news: Fundamental theo-
ries, detection methods, and opportunities. ACM Computing Surveys
(CSUR) 53(5), 1–40 (2020)
[23] MacDougall, C.D.: Hoaxes vol. 465. Dover Publications, ??? (1958)
[24] Kapferer, J.-N.: Rumors Uses, Interpretations, & Images. Routledge, ???
(2017)
[25] Alzanin, S.M., Azmi, A.M.: Detecting rumors in social media: A survey.
Procedia computer science 142, 294–300 (2018)
[26] Rubin, V.L., Chen, Y., Conroy, N.K.: Deception detection for news: three
types of fakes. Proceedings of the Association for Information Science
and Technology 52(1), 1–4 (2015)
[27] Woolley, S.C., Howard, P.N.: Computational Propaganda: Political Par-
ties, Politicians, and Political Manipulation on Social Media. Oxford
University Press, ??? (2018)
[28] Volkova, S., Shaffer, K., Jang, J.Y., Hodas, N.: Separating facts from
fiction: Linguistic models to classify suspicious and trusted news posts on
twitter. In: Proceedings of the 55th Annual Meeting of the Association
for Computational Linguistics (volume 2: Short Papers), pp. 647–653
(2017)
[29] Chen, Y., Conroy, N.J., Rubin, V.L.: Misleading online content: recog-
nizing clickbait as” false news”. In: Proceedings of the 2015 ACM on
Workshop on Multimodal Deception Detection, pp. 15–19 (2015)
[30] Fetzer, J.H.: Disinformation: The use of false information. Minds and
Machines 14(2), 231–240 (2004)

40 A Comparative Study of Computational Fake News Detection on Social Media
[31] Kucharski, A.: Study epidemiology of fake news. Nature 540(7634), 525–
525 (2016)
[32] Fetzer, J.H.: Information: Does it have to be true? Minds and Machines
14(2), 223–229 (2004)
[33] Kshetri, N., Voas, J.: The economics of “fake news”. IT Professional
19(6), 8–12 (2017)
[34] Wardle, C., Derakhshan, H., et al.: Thinking about ‘information dis-
order’: formats of misinformation, disinformation, and mal-information.
Ireton, Cherilyn; Posetti, Julie. Journalism,‘fake news’& disinformation.
Paris: Unesco, 43–54 (2018)
[35] Elhadad, M.K., Li, K.F., Gebali, F.: Fake news detection on social media:
a systematic survey. In: 2019 IEEE Pacific Rim Conference on Commu-
nications, Computers and Signal Processing (PACRIM), pp. 1–8 (2019).
IEEE
[36] Howard, P.N., Kollanyi, B.: Bots,# strongerin, and# brexit: compu-
tational propaganda during the uk-eu referendum. Available at SSRN
2798311 (2016)
[37] Ferrara, E.: Disinformation and social bot operations in the run up to
the 2017 french presidential election. arXiv preprint arXiv:1707.00086
(2017)
[38] Rapoza, K.: Can ‘fake news’ impact the stock market? Forbes News
(2017)
[39] Ferrara, E., Varol, O., Davis, C., Menczer, F., Flammini, A.: The rise of
social bots. Communications of the ACM 59(7), 96–104 (2016)
[40] How fake news in the middle east is a powder keg waiting to blow.
Independent (September 2017)
[41] Spohr, D.: Fake news and ideological polarization: Filter bubbles and
selective exposure on social media. Business Information Review 34(3),
150–160 (2017)
[42] Shayan Sardarizadeh, O.R.: Ukraine invasion: False claims the war is a
hoax go viral. BBC Monitoring (March 2022)
[43] Salve, P.: Manipulative fake news on the rise in india under lockdown:
Study. indiaspend.com (May 2020)
[44] PTI: Govt orders blocking of 18 indian, four pakistani youtube-based
news channels. timesofindia.indiatimes.com (April 2022)

A Comparative Study of Computational Fake News Detection on Social Media 41
[45] Zhang, X., Ghorbani, A.A.: An overview of online fake news: Characteri-
zation, detection, and discussion. Information Processing & Management
57(2), 102025 (2020)
[46] Liu, Y., Wu, Y.-F.: Early detection of fake news on social media through
propagation path classification with recurrent and convolutional net-
works. In: Proceedings of the AAAI Conference on Artificial Intelligence,
vol. 32 (2018)
[47] Shao, C., Hui, P.-M., Wang, L., Jiang, X., Flammini, A., Menczer, F.,
Ciampaglia, G.L.: Anatomy of an online misinformation network. Plos
one 13(4), 0196087 (2018)
[48] Gentzkow, M., Shapiro, J.M., Stone, D.F.: Media bias in the market-
place: Theory. In: Handbook of Media Economics vol. 1, pp. 623–645.
Elsevier, ??? (2015)
[49] Wang, Y., Ma, F., Jin, Z., Yuan, Y., Xun, G., Jha, K., Su, L.,
Gao, J.: Eann: Event adversarial neural networks for multi-modal fake
news detection. In: Proceedings of the 24th Acm Sigkdd International
Conference on Knowledge Discovery & Data Mining, pp. 849–857 (2018)
[50] Shu, K., Wang, S., Liu, H.: Beyond news contents: The role of social
context for fake news detection. In: Proceedings of the Twelfth ACM
International Conference on Web Search and Data Mining, pp. 312–320
(2019)
[51] Wu, Z., Pan, S., Chen, F., Long, G., Zhang, C., Philip, S.Y.: A compre-
hensive survey on graph neural networks. IEEE transactions on neural
networks and learning systems 32(1), 4–24 (2020)
[52] Wu, K., Yang, S., Zhu, K.Q.: False rumors detection on sina weibo by
propagation structures. In: 2015 IEEE 31st International Conference on
Data Engineering, pp. 651–662 (2015). IEEE
[53] F¨urnkranz, J.: A study using n-gram features for text categoriza-
tion. Austrian Research Institute for Artifical Intelligence 3(1998), 1–10
(1998)
[54] Jin, Z., Cao, J., Zhang, Y., Zhou, J., Tian, Q.: Novel visual and statistical
image features for microblogs news verification. IEEE transactions on
multimedia 19(3), 598–608 (2016)
[55] Kwon, S., Cha, M., Jung, K., Chen, W., Wang, Y.: Prominent fea-
tures of rumor propagation in online social media. In: 2013 IEEE 13th
International Conference on Data Mining, pp. 1103–1108 (2013). IEEE

42 A Comparative Study of Computational Fake News Detection on Social Media
[56] Ma, J., Gao, W., Wei, Z., Lu, Y., Wong, K.-F.: Detect rumors using time
series of social context information on microblogging websites. In: Pro-
ceedings of the 24th ACM International on Conference on Information
and Knowledge Management, pp. 1751–1754 (2015)
[57] Ruchansky, N., Seo, S., Liu, Y.: Csi: A hybrid deep model for fake
news detection. In: Proceedings of the 2017 ACM on Conference on
Information and Knowledge Management, pp. 797–806 (2017)
[58] Jin, Z., Cao, J., Zhang, Y., Luo, J.: News verification by exploiting con-
flicting social viewpoints in microblogs. In: Proceedings of the AAAI
Conference on Artificial Intelligence, vol. 30 (2016)
[59] Potthast, M., Kiesel, J., Reinartz, K., Bevendorff, J., Stein, B.: A
stylometric inquiry into hyperpartisan and fake news. arXiv preprint
arXiv:1702.05638 (2017)
[60] Tacchini, E., Ballarin, G., Della Vedova, M.L., Moret, S., de Alfaro, L.:
Some like it hoax: Automated fake news detection in social networks.
arXiv preprint arXiv:1704.07506 (2017)
[61] Assiroj, P., Hidayanto, A.N., Prabowo, H., Warnars, H.L.H.S., et al.:
Hoax news detection on social media: A survey. In: 2018 Indonesian
Association for Pattern Recognition International Conference (INAPR),
pp. 186–191 (2018). IEEE
[62] Vlachos, A., Riedel, S.: Fact checking: Task definition and dataset con-
struction. In: Proceedings of the ACL 2014 Workshop on Language
Technologies and Computational Social Science, pp. 18–22 (2014)
[63] Cohen, S., Hamilton, J.T., Turner, F.: Computational journalism. Com-
munications of the ACM 54(10), 66–71 (2011)
[64] Wang, W.Y.: ” liar, liar pants on fire”: A new benchmark dataset for
fake news detection. arXiv preprint arXiv:1705.00648 (2017)
[65] Rubin, V.L., Lukoianova, T.: Truth and deception at the rhetorical
structure level. Journal of the Association for Information Science and
Technology 66(5), 905–917 (2015)
[66] Atanasova, P., Nakov, P., M`arquez, L., Barr´on-Cede˜no, A., Karadzhov,
G., Mihaylova, T., Mohtarami, M., Glass, J.: Automatic fact-checking
using context and discourse information. Journal of Data and Informa-
tion Quality (JDIQ) 11(3), 1–27 (2019)
[67] Mohammad, S.M., Sobhani, P., Kiritchenko, S.: Stance and sentiment in
tweets. ACM Transactions on Internet Technology (TOIT) 17(3), 1–23

A Comparative Study of Computational Fake News Detection on Social Media 43
(2017)
[68] Gupta, M., Zhao, P., Han, J.: Evaluating event credibility on twitter. In:
Proceedings of the 2012 SIAM International Conference on Data Mining,
pp. 153–164 (2012). SIAM
[69] Jin, Z., Cao, J., Jiang, Y.-G., Zhang, Y.: News credibility evaluation
on microblog with a hierarchical propagation model. In: 2014 IEEE
International Conference on Data Mining, pp. 230–239 (2014). IEEE
[70] Webster, J.J., Kit, C.: Tokenization as the initial phase in nlp. In:
COLING 1992 Volume 4: The 14th International Conference on Com-
putational Linguistics (1992)
[71] Lovins, J.B.: Development of a stemming algorithm. Mech. Transl.
Comput. Linguistics 11(1-2), 22–31 (1968)
[72] Zhang, Y., Jin, R., Zhou, Z.-H.: Understanding bag-of-words model: a
statistical framework. International Journal of Machine Learning and
Cybernetics 1(1), 43–52 (2010)
[73] Cavnar, W.B., Trenkle, J.M., et al.: N-gram-based text categorization.
In: Proceedings of SDAIR-94, 3rd Annual Symposium on Document
Analysis and Information Retrieval, vol. 161175 (1994). Citeseer
[74] Ramos, J., et al.: Using tf-idf to determine word relevance in document
queries. In: Proceedings of the First Instructional Conference on Machine
Learning, vol. 242, pp. 29–48 (2003). Citeseer
[75] Fransiska, S., Rianto, R., Gufroni, A.I.: Sentiment analysis provider by.
u on google play store reviews with tf-idf and support vector machine
(svm) method. Scientific Journal of Informatics 7(2), 203–212 (2020)
[76] Teslya, N., Savosin, S.: Matching ontologies with word2vec-based neural
network. In: International Conference on Computational Science and Its
Applications, pp. 745–756 (2019). Springer
[77] Kumawat, D., Jain, V.: Pos tagging approaches: A comparison. Interna-
tional Journal of Computer Applications 118(6) (2015)
[78] Moher, D., Liberati, A., Tetzlaff, J., Altman, D.G., Altman, D., Antes,
G., Atkins, D., Barbour, V., Barrowman, N., Berlin, J.A., et al.:
Preferred reporting items for systematic reviews and meta-analyses:
the prisma statement (chinese edition). Journal of Chinese Integrative
Medicine 7(9), 889–896 (2009)
[79] Oshikawa, R., Qian, J., Wang, W.Y.: A survey on natural language

44 A Comparative Study of Computational Fake News Detection on Social Media
processing for fake news detection. arXiv preprint arXiv:1811.00770
(2018)
[80] Shu, K., Wang, S., Liu, H.: Understanding user profiles on social
media for fake news detection. In: 2018 IEEE Conference on Multime-
dia Information Processing and Retrieval (MIPR), pp. 430–435 (2018).
IEEE
[81] Shu, K., Mahudeswaran, D., Liu, H.: Fakenewstracker: a tool for
fake news collection, detection, and visualization. Computational and
Mathematical Organization Theory 25(1), 60–71 (2019)
[82] Yang, S., Shu, K., Wang, S., Gu, R., Wu, F., Liu, H.: Unsupervised fake
news detection on social media: A generative approach. In: Proceedings
of the AAAI Conference on Artificial Intelligence, vol. 33, pp. 5644–5651
(2019)
[83] Boididou, C., Andreadou, K., Papadopoulos, S., Dang-Nguyen, D.-T.,
Boato, G., Riegler, M., Kompatsiaris, Y., et al.: Verifying multimedia
use at mediaeval 2015. MediaEval 3(3), 7 (2015)
[84] Sharma, K., Qian, F., Jiang, H., Ruchansky, N., Zhang, M., Liu,
Y.: Combating fake news: A survey on identification and mitigation
techniques. ACM Transactions on Intelligent Systems and Technology
(TIST) 10(3), 1–42 (2019)
[85] Meyn, L.F.: Fake news prediction on facebook: Design and implementa-
tion of a fake news prediction tool. PhD thesis, University of Nebraska
at Omaha (2018)
[86] Zhang, X., Cao, J., Li, X., Sheng, Q., Zhong, L., Shu, K.: Mining dual
emotion for fake news detection. In: Proceedings of the Web Conference
2021, pp. 3465–3476 (2021)
[87] Cardoso Durier da Silva, F., Vieira, R., Garcia, A.C.: Can machines learn
to detect fake news? a survey focused on social media. In: Proceedings
of the 52nd Hawaii International Conference on System Sciences (2019)
[88] Pan, J.Z., Pavlova, S., Li, C., Li, N., Li, Y., Liu, J.: Content based fake
news detection using knowledge graphs. In: International Semantic Web
Conference, pp. 669–683 (2018). Springer
[89] Bharadwaj, P., Shao, Z.: Fake news detection with semantic features
and text mining. International Journal on Natural Language Computing
(IJNLC) Vol 8(2019)
[90] Buntain, C., Golbeck, J.: Automatically identifying fake news in popular

A Comparative Study of Computational Fake News Detection on Social Media 45
twitter threads. In: 2017 IEEE International Conference on Smart Cloud
(SmartCloud), pp. 208–215 (2017). IEEE
[91] Parikh, S.B., Atrey, P.K.: Media-rich fake news detection: A survey.
In: 2018 IEEE Conference on Multimedia Information Processing and
Retrieval (MIPR), pp. 436–441 (2018). IEEE
[92] Ajao, O., Bhowmik, D., Zargari, S.: Fake news identification on twitter
with hybrid cnn and rnn models. In: Proceedings of the 9th International
Conference on Social Media and Society, pp. 226–230 (2018)
[93] Ni, S., Li, J., Kao, H.-Y.: Mvan: Multi-view attention networks for fake
news detection on social media. IEEE Access 9, 106907–106917 (2021)
[94] Mitra, T., Gilbert, E.: Credbank: A large-scale social media corpus with
associated credibility annotations. In: Proceedings of the International
AAAI Conference on Web and Social Media, vol. 9, pp. 258–267 (2015)
[95] Lu, Y.-J., Li, C.-T.: Gcan: Graph-aware co-attention networks for
explainable fake news detection on social media. arXiv preprint
arXiv:2004.11648 (2020)
[96] Della Vedova, M.L., Tacchini, E., Moret, S., Ballarin, G., DiPierro, M., de
Alfaro, L.: Automatic online fake news detection combining content and
social signals. In: 2018 22nd Conference of Open Innovations Association
(FRUCT), pp. 272–279 (2018). IEEE
[97] Gupta, S., Thirukovalluru, R., Sinha, M., Mannarswamy, S.: Cimtdetect:
a community infused matrix-tensor coupled factorization based method
for fake news detection. In: 2018 IEEE/ACM International Conference
on Advances in Social Networks Analysis and Mining (ASONAM), pp.
278–281 (2018). IEEE
[98] Zhang, J., Dong, B., Philip, S.Y.: Fakedetector: Effective fake news detec-
tion with deep diffusive neural network. In: 2020 IEEE 36th International
Conference on Data Engineering (ICDE), pp. 1826–1829 (2020). IEEE
[99] Shu, K., Mahudeswaran, D., Wang, S., Lee, D., Liu, H.: Fakenewsnet:
A data repository with news content, social context and spatialtempo-
ral information for studying fake news on social media. arXiv preprint
arXiv:1809.01286 (2018)
[100] Shu, K., Zhou, X., Wang, S., Zafarani, R., Liu, H.: The role of user
profiles for fake news detection. In: Proceedings of the 2019 IEEE/ACM
International Conference on Advances in Social Networks Analysis and
Mining, pp. 436–439 (2019)

46 A Comparative Study of Computational Fake News Detection on Social Media
[101] Shu, K., Mahudeswaran, D., Wang, S., Liu, H.: Hierarchical propaga-
tion networks for fake news detection: Investigation and exploitation. In:
Proceedings of the International AAAI Conference on Web and Social
Media, vol. 14, pp. 626–637 (2020)
[102] Zhou, X., Jain, A., Phoha, V.V., Zafarani, R.: Fake news early detection:
An interdisciplinary study. arXiv preprint arXiv:1904.11679 (2019)
[103] Kotteti, C.M.M., Dong, X., Li, N., Qian, L.: Fake news detection
enhancement with data imputation. In: 2018 IEEE 16th Intl Conf
on Dependable, Autonomic and Secure Computing, 16th Intl Conf on
Pervasive Intelligence and Computing, 4th Intl Conf on Big Data Intel-
ligence and Computing and Cyber Science and Technology Congress
(DASC/PiCom/DataCom/CyberSciTech), pp. 187–192 (2018). IEEE
[104] Fern´andez-Reyes, F.C., Shinde, S.: Evaluating deep neural networks for
automatic fake news detection in political domain. In: Ibero-American
Conference on Artificial Intelligence, pp. 206–216 (2018). Springer
[105] Dong, M., Yao, L., Wang, X., Benatallah, B., Sheng, Q.Z., Huang, H.:
Dual: A deep unified attention model with latent relation representations
for fake news detection. In: International Conference on Web Information
Systems Engineering, pp. 199–209 (2018). Springer
[106] Karimi, H., Roy, P., Saba-Sadiya, S., Tang, J.: Multi-source multi-class
fake news detection. In: Proceedings of the 27th International Conference
on Computational Linguistics, pp. 1546–1557 (2018)
[107] Olivieri, A., Shabani, S., Sokhn, M., Cudr´e-Mauroux, P.: Creating task-
generic features for fake news detection. In: Proceedings of the 52nd
Hawaii International Conference on System Sciences (2019)
[108] Rasool, T., Butt, W.H., Shaukat, A., Akram, M.U.: Multi-label fake
news detection using multi-layered supervised learning. In: Proceedings
of the 2019 11th International Conference on Computer and Automation
Engineering, pp. 73–77 (2019)
[109] Janze, C., Risius, M.: Automatic detection of fake news on social media
platforms (2017)
[110] Reis, J.C., Correia, A., Murai, F., Veloso, A., Benevenuto, F.: Supervised
learning for fake news detection. IEEE Intelligent Systems 34(2), 76–81
(2019)
[111] Zubiaga, A., Liakata, M., Procter, R., Wong Sak Hoi, G., Tolmie, P.:
Analysing how people orient to and spread rumours in social media by
looking at conversational threads. PloS one 11(3), 0150989 (2016)

A Comparative Study of Computational Fake News Detection on Social Media 47
[112] Zubiaga, A., Liakata, M., Procter, R.: Exploiting context for rumour
detection in social media. In: International Conference on Social Infor-
matics, pp. 109–123 (2017). Springer
[113] Jin, Z., Cao, J., Guo, H., Zhang, Y., Luo, J.: Multimodal fusion with
recurrent neural networks for rumor detection on microblogs. In: Pro-
ceedings of the 25th ACM International Conference on Multimedia, pp.
795–816 (2017)
[114] Yuan, C., Ma, Q., Zhou, W., Han, J., Hu, S.: Jointly embedding the local
and global relations of heterogeneous graph for rumor detection. In: The
19th IEEE International Conference on Data Mining (2019). IEEE
[115] Min, E., Rong, Y., Bian, Y., Xu, T., Zhao, P., Huang, J., Ananiadou, S.:
Divide-and-conquer: Post-user interaction network for fake news detec-
tion on social media. In: Proceedings of the ACM Web Conference 2022,
pp. 1148–1158 (2022)
[116] Rubin, V.L., Conroy, N.J., Chen, Y.: Towards news verification: Decep-
tion detection methods for news discourse. In: Hawaii International
Conference on System Sciences, pp. 5–8 (2015)
[117] Pennebaker, J.W., Boyd, R.L., Jordan, K., Blackburn, K.: The devel-
opment and psychometric properties of liwc2015. Technical report
(2015)
[118] Ferreira, W., Vlachos, A.: Emergent: a novel data-set for stance classi-
fication. In: Proceedings of the 2016 Conference of the North American
Chapter of the Association for Computational Linguistics: Human
Language Technologies (2016). ACL
[119] ¨
Ozg¨obek, ¨
O., Gulla, J.A.: Towards an understanding of fake news. In:
CEUR Workshop Proceedings, vol. 354 (2017)
[120] Shu, K., Wang, S., Liu, H.: Exploiting tri-relationship for fake news
detection. arXiv preprint arXiv:1712.07709 8(2017)
[121] Hanselowski, A., PVS, A., Schiller, B., Caspelherr, F., Chaudhuri, D.,
Meyer, C.M., Gurevych, I.: A retrospective analysis of the fake news
challenge stance detection task. arXiv preprint arXiv:1806.05180 (2018)
[122] Atodiresei, C.-S., T˘an˘aselea, A., Iftene, A.: Identifying fake news and
fake users on twitter. Procedia Computer Science 126, 451–461 (2018)
[123] Meinert, J., Mirbabaie, M., Dungs, S., Aker, A.: Is it really fake?–towards
an understanding of fake news in social media communication. In: Inter-
national Conference on Social Computing and Social Media, pp. 484–497

48 A Comparative Study of Computational Fake News Detection on Social Media
(2018). Springer
[124] Kumar, S., Shah, N.: False information on web and social media: A
survey. arXiv preprint arXiv:1804.08559 (2018)
[125] de Alfaro, L., Di Pierro, M., Agrawal, R., Tacchini, E., Ballarin, G.,
Della Vedova, M.L., Moret, S.: Reputation systems for news on twitter:
A large-scale study. arXiv preprint arXiv:1802.08066 (2018)
[126] Zhang, J., Cui, L., Fu, Y., Gouza, F.B.: Fake news detection with deep
diffusive network model. arXiv preprint arXiv:1805.08751 (2018)
[127] Wu, L., Liu, H.: Tracing fake-news footprints: Characterizing social
media messages by how they propagate. In: Proceedings of the Eleventh
ACM International Conference on Web Search and Data Mining, pp.
637–645 (2018)
[128] Guacho, G.B., Abdali, S., Shah, N., Papalexakis, E.E.: Semi-supervised
content-based detection of misinformation via tensor embeddings. In:
2018 IEEE/ACM International Conference on Advances in Social Net-
works Analysis and Mining (ASONAM), pp. 322–325 (2018). IEEE
[129] Helmstetter, S., Paulheim, H.: Weakly supervised learning for fake news
detection on twitter. In: 2018 IEEE/ACM International Conference on
Advances in Social Networks Analysis and Mining (ASONAM), pp. 274–
277 (2018). IEEE
[130] Guo, C., Cao, J., Zhang, X., Shu, K., Yu, M.: Exploiting emotions for
fake news detection on social media. arXiv preprint arXiv:1903.01728
(2019)
[131] Traylor, T., Straub, J., Snell, N., et al.: Classifying fake news arti-
cles using natural language processing to identify in-article attribution
as a supervised learning estimator. In: 2019 IEEE 13th International
Conference on Semantic Computing (ICSC), pp. 445–449 (2019). IEEE
[132] Zhou, X., Jain, A., Phoha, V.V., Zafarani, R.: Fake news early detection:
A theory-driven model. Digital Threats: Research and Practice 1(2),
1–25 (2020)
[133] Kesarwani, A., Chauhan, S.S., Nair, A.R.: Fake news detection on social
media using k-nearest neighbor classifier. In: 2020 International Con-
ference on Advances in Computing and Communication Engineering
(ICACCE), pp. 1–4 (2020). IEEE
[134] Gimpel, H., Heger, S., Kasper, J., Sch¨afer, R.: The power of related
articles–improving fake news detection on social media platforms (2020)

A Comparative Study of Computational Fake News Detection on Social Media 49
[135] Wang, Y., Zhang, Y., Li, X., Yu, X.: Covid-19 fake news detection using
bidirectional encoder representations from transformers based models.
arXiv preprint arXiv:2109.14816 (2021)
[136] Sahoo, S.R., Gupta, B.B.: Multiple features based approach for auto-
matic fake news detection on social networks using deep learning.
Applied Soft Computing 100, 106983 (2021)
[137] Jarrahi, A., Safari, L.: Fr-detect: A multi-modal framework for early fake
news detection on social media using publishers features. arXiv preprint
arXiv:2109.04835 (2021)
[138] Shu, K., Mosallanezhad, A., Liu, H.: Cross-domain fake news detection
on social media: A context-aware adversarial approach. In: Frontiers in
Fake Media Generation and Detection, pp. 215–232. Springer, ??? (2022)
[139] Seddari, N., Derhab, A., Belaoued, M., Halboob, W., Al-Muhtadi, J.,
Bouras, A.: A hybrid linguistic and knowledge-based analysis approach
for fake news detection on social media. IEEE Access (2022)
[140] Qazvinian, V., Rosengren, E., Radev, D., Mei, Q.: Rumor has it: Identify-
ing misinformation in microblogs. In: Proceedings of the 2011 Conference
on Empirical Methods in Natural Language Processing, pp. 1589–1599
(2011)
[141] Popat, K.: Assessing the credibility of claims on the web. In: Proceedings
of the 26th International Conference on World Wide Web Companion,
pp. 735–739 (2017)
[142] Jin, F., Dougherty, E., Saraf, P., Cao, Y., Ramakrishnan, N.: Epidemi-
ological modeling of news and rumors on twitter. In: Proceedings of the
7th Workshop on Social Network Mining and Analysis, pp. 1–9 (2013)
[143] Sampson, J., Morstatter, F., Wu, L., Liu, H.: Leveraging the implicit
structure within social media for emergent rumor detection. In: Proceed-
ings of the 25th ACM International on Conference on Information and
Knowledge Management, pp. 2377–2382 (2016)
[144] Ma, J., Gao, W., Wong, K.-F.: Detect rumors in microblog posts
using propagation structure via kernel learning. (2017). Association for
Computational Linguistics
[145] Silverman, C.: This analysis shows how viral fake election news stories
outperformed real news on facebook. BuzzFeed news 16 (2016)
[146] Castillo, C., Mendoza, M., Poblete, B.: Predicting information credibility
in time-sensitive social media. Internet Research (2013)

50 A Comparative Study of Computational Fake News Detection on Social Media
[147] Zubiaga, A., Liakata, M., Procter, R.: Learning reporting dynamics dur-
ing breaking news for rumour detection in social media. arXiv preprint
arXiv:1610.07363 (2016)