ArticlePDF Available

A TURKISH MAKAM MUSIC SYMBOLIC DATABASE FOR MUSIC INFORMATION RETRIEVAL: SymbTr

Authors:

Abstract and Figures

Turkish makam music needs a comprehensive database for public consumption, to be used in MIR. This article introduces SymbTr, a Turkish Makam Music Symbolic Representation Database, aimed at filling this void. SymbTr consists of musical information in text, PDF, and MIDI formats. Raw data, drawn from reliable sources, and consisting of 1,700 musical pieces in Turkish art and folk music was processed featuring distinct examples in 155 diverse makams, 100 usuls and 48 forms. Special care was devoted to selection of works that scatter across a broad historical time span and were among those still performed today. Total number of musical notes in these pieces was 630,000, corresponding to a nominal playback time of 72 hours. Synthesized sounds particular to Turk-ish makam music were used in MIDI playback, and tran-scription/playback errors were corrected by input from experts. Symbolic representation data, open to the public, is output from a computer program developed exclusively for Turkish makam music. SymbTr was designed as a wholesome representation of aforementioned distinct au-ditory and visual features that distinguish Turkish makam music from other music genres. This article explains the database format in detail, and also provides, through ex-amples, statistical information on pitch/interval allocation and distribution.
Content may be subject to copyright.
A TURKISH MAKAM MUSIC SYMBOLIC DATABASE FOR
MUSIC INFORMATION RETRIEVAL: SymbTr
M. Kemal Karaosmanoğlu
Yıldız Technical University
kkara@yildiz.edu.tr
ABSTRACT
Turkish makam music needs a comprehensive database
for public consumption, to be used in MIR. This article
introduces SymbTr, a Turkish Makam Music Symbolic
Representation Database, aimed at filling this void.
SymbTr consists of musical information in text, PDF, and
MIDI formats. Raw data, drawn from reliable sources,
and consisting of 1,700 musical pieces in Turkish art and
folk music was processed featuring distinct examples in
155 diverse makams, 100 usuls and 48 forms. Special
care was devoted to selection of works that scatter across
a broad historical time span and were among those still
performed today. Total number of musical notes in these
pieces was 630,000, corresponding to a nominal playback
time of 72 hours. Synthesized sounds particular to Turk-
ish makam music were used in MIDI playback, and tran-
scription/playback errors were corrected by input from
experts. Symbolic representation data, open to the public,
is output from a computer program developed exclusively
for Turkish makam music. SymbTr was designed as a
wholesome representation of aforementioned distinct au-
ditory and visual features that distinguish Turkish makam
music from other music genres. This article explains the
database format in detail, and also provides, through ex-
amples, statistical information on pitch/interval allocation
and distribution.
1. INTRODUCTION
Turkish makam music is a genre drawing roots from a
thousand year old tradition, featuring distinct melodic
patterns called makam and rich rhythmic structures called
usul. Since the number of tones per octave is greater in
Turkish makam music, compared to Western music, sev-
eral sharp and flat accidentals appear in printed scores.
Additionally, one must take into consideration a multi-
tude of idiosyncratic rhythmic structures. Although there
exists only one version of the score, independent of the
instrument or key, musicians perform improvised trans-
positions during performance, as permitted by the ranges
of their instruments and the vocalist on hand. Probably
the most prominent feature of Turkish makam music is its
monophonic ─and incidentally heterophonic─ structure.
Another characteristic is the number of notes in an oc-
tave: 17, 24, and, according to some musicologists, even
a greater number of tones to the octave make up the pitch
palette of Turkish makam music [12], [16]. Although dis-
playing a higher pitch count compared to Western music,
there is no one-to-one correlation between the fixed fre-
quency values, music theory, implied in engraved scores
and what is actually performed in practice [3].
Everything mentioned up to this point was to differ-
entiate Turkish makam music from many other world
music genres. It then follows; data structures and algo-
rithms developed for other musical traditions are not di-
rectly applicable to Turkish makam music. On the other
hand, there are only a handful of researchers working on
computational models for Turkish makam music. There
remains much to be done in areas related to data collec-
tion/compilation, algorithm development, and research.
SymbTr is hopefully a likely candidate to be a pioneer in
the field, since it is capable of accommodating and ex-
pressing information specific to makam music. Secondly,
early studies ([9], [18]) have returned encouraging re-
sults. It is anticipated that SymbTr might provide a setting
for scholars interested in makam music, potentially
(a)
(b)
(c)
Figure 1. A Turkish folksong's scoring in (a) KTM,
(b) THM, and (c) mixed format
Permission to make digital or hard copies of all or part of this work for
personal or classroom use is granted without fee provided that copies
are not made or distributed for profit or commercial advantage and that
copies bear this notice and the full citation on the first page.
© 2012 International Society for Music Information Retrieval
stimulating further research at a global level.
2. STYLES OF TURKISH MAKAM MUSIC
Turkish makam music is viewed under two major head-
ings:
1. Classical Turkish music (KTM),
2. Turkish folk music (THM).
Since both styles originate from the same cultural
roots, their modal motifs and rhythmic structures are very
similar in character [17]. Owing to political movements
emerging at the turn of the 20th Century, a superficial bi-
furcation took place, which led to a divergence between
the two styles resulting in two separate traditions. Today,
these two traditions differ considerably when it comes to
their respective theoretical models, notation systems, and
terminology. Fig. 1 shows the first two measures from a
folksong score, which is part of both the THM and KTM
repertoires of TRT
1
. Scores are shown in three different
notational systems.
As can be detected in the scores, accidental symbols,
in particular, are different even though the melody is es-
sentially the same: In the KTM version reverse and
hooked flat signs (Fig. 1-a) represent the accidentals,
while in the THM version superscripts over ordinary flat
signs are used (Fig. 1-b). The mixed notation in (Fig. 1-c)
displays a combination of the two. Reverse and hooked
flats, by definition, lower a note by 1 and 4 Holdrian
commas (Hc)
2
respectively. However, many of the
measurements ([1], [4], [5]) evince that, printed scores for
works in the Saba makam should carry a 2 comma flat
sign for B and a 3 comma flat sign for D as the key signa-
ture. Indeed, these values are substituted in THM and
mixed notations.
The difference between KTM and THM notation lies
not only in the symbols representing the accidentals. In
KTM notation there are almost no ornamentation sym-
bols on the score. In THM, on the other hand, ornamenta-
tion is achieved by repeated use of notes with smaller
rhythmic values, as shown in the trill in Fig. 1-b [17].
When such passages are converted into SymbTr, note
clusters representing ornamentation are indicated by a
single core note, with its type shown in the Code field.
Because of the fact that THM and KTM have mixed
and intertwined traditionally, the SymbTr database natu-
rally accommodates pieces from both categories. Format
in the database was, therefore, designed to reconcile the
artificial disparity between the two traditions. The most
important design element for the database was the fun-
damental tuning selected. The Arel - Ezgi (AE) tone-
system, which has been recognized and widely adopted as
the official KTM system since the 1950s, has 24 notes in
an octave. In contrast, THM has adopted a notation with
17 notes to the octave. Twelve of the said 17 notes are
common with the AE system. Moreover, both tonal scales
1
Turkish Radio/TV Broadcasting Corporation.
2
Interval unit obtained by the division of the octave into logarithmical-
ly equal 53 parts: Hc = 1200 / 53 ≈ 22.5 cent. In this article comma
signifies the Holdrian comma.
present a near-perfect subset of 53 tone equal tempera-
ment (53TET), with deviations less than 1 cent [19] (Fig.
2). Possibly due to this structural connection, Turkish
makam music education has been built around 53TET,
whether acknowledged by name (Ayomak, Sarısözen) or
Figure 2. 17 tones in THM (left column), 24 tones in
KTM (right column), and 53TET in between
not [13]. Hence, the term "comma", when describing
makams and preparing printed scores, refers to the
Holdrian comma as the basic intervallic unit, obtained by
equally dividing the octave into 53 equal parts. Selecting
53TET as the master underlying tuning in SymbTr also
facilitates transpositions across ahenks (pitch-levels).
Ahenks can be defined as 7 principal and 5 minor catego-
ries corresponding to 12 chromatic pitch-levels akin to
what key transposing instruments of Western music ac-
complish. Detailed information about ahenks and Turkish
makam music in general can be found in [10] and [14].
3. MAKAM MUSIC AND SYMBOLIC DATA
SymbTr database is generated by using the output from a
computer program Mus2-Alpha, developed by the author
of this article. This software is the first notation and play-
back application for Turkish makam music to the best of
our knowledge. All pieces in the database were entered
manually using the said software. Printed scores and
MIDI files were, then, prepared for every piece in the da-
tabase. Initially, before the introduction of Mus2-Alpha
and its sister applications (Nota 2.2
1
, Notist
2
), scores
were engraved either manually or using programs such as
Finale or Sibelius, that were developed solely to tran-
scribe Western music. Since these programs were not de-
signed to notate flats and sharps specific to Turkish
makam music, their standard output formats such as Mu-
sicXML and MIDI have not been useful in research on
Turkish makam music [7].
The format for SymbTr described in this article was
derived from Mus2-Alpha's original format that was used
initially to transcribe printable sheet music for pieces in
Turkish makam music. Since this format includes reprise
markings such as segno and coda, some modifications for
scientific research are necessary. In SymbTr, notes are
linearized just as they are performed. An advantage asso-
ciated with Mus2-Alpha originating data is that pieces can
be amended through consultation with experts, using lis-
tening tests based on synthesized sound output. An entry
level version of this program, Mus2okur
3
, has reached
thousands of users, thereby resulting in a wide scale
screening of possible errors in the database.
The main source of data in SymbTr is TRT and other
trustworthy archives (Recollection of Turkish Music Cul-
ture
4
), where almost all of them were entered using the
AE notation. To synthesize realistic intonations, however,
it was necessary to use pitches not included in the AE
tone-system. Five notes in the THM scale lie outside the
AE scale (Fig. 2). As a courtesy for Turkish musicians, a
composite system was adopted in the printout scores of
SymbTr: Symbols for flats and sharps were taken directly
from AE, and numerical superscripts were inserted to ex-
press comma-alterations for notes that were not available
1
http://www.tulgan.com/Nota22/
2
http://notist.org/
3
www.musiki.org
4
www.sanatmuziginotalari.com/ under http://devletkorosu.com
Please go to the second link ‘http’first to reach the main site. Then,
look for and click on the first address ‘www’.
in the tone-system (Fig. 1-c).
4. SymbTr FORMAT
Basic information such as makam, form and usul related
to each piece in SymbTr is indicated in the filename. In
this manner, any piece can be accessed directly from the
file system:
beyati--sarki--aksak--karsidan_yar--dede_efendi.txt
Makam Form Usul Title Composer
Some fields in the SymbTr format consist of different
representations of the same information. Therefore, one
field can be easily converted into the other with the help
of the relevant computer code. However, since this addi-
tional information requires very little extra storage space,
it is provided separately for the convenience of research-
ers. These basic and readily derived fields are described
under common headings below.
Code: Signifies a normal note (#9) or ornamentation.
The most commonly used ornamentation codes are as fol-
lows: #7 for tremolos, #8 for acciaccatura, #12 for trills,
and #23 for mordent.
NoteAE / CommaAE: A kind of scientific pitch nota-
tion [20]: Indicates note letter, its octave (for exam-
ple, G5 for gerdaniye), and its comma equivalent (349)
(Fig. 3). Notes in THM sheets that do not exist in the AE
system are represented by their closest equivalent
AE note, e.g. Mi b2 = Dikhisar (Eb1) (Fig. 2). C4 is the
Code
Note53
Comma53
NoteAE
CommaAE
Num.
Denom.
ms
LNS
VelOn
Syllable
9
Do5
318
C5
318
1
4
667
95
96
Bir
9
Re5b3
324
D5b4
325
1
8
333
99
108
dal
9
Re5b3
324
D5b4
325
1
16
167
99
96
9
Do5
318
C5
318
1
16
167
95
84
9
Si4b2
312
B4b1
313
1
4
667
95
72
da
12
Do5
318
C5
318
1
8
333
99
96
i
9
Do5
318
C5
318
1
16
167
99
96
9
Si4b2
312
B4b1
313
1
16
167
95
96
9
La4
305
A4
305
1
8
333
99
96
ki
8
Si4b2
312
B4b1
313
1
8
42
99
96
9
La4
305
A4
305
1
16
167
99
96
9
Sol4
296
G4
296
1
16
167
95
96
9
La4
305
A4
305
1
8
333
99
96
ki
9
Si4b2
312
B4b1
313
1
8
333
95
96
9
Do5
318
C5
318
1
4
1334
45
84
raz
Table 1. SymbTr representation of the score in Fig. 1.c
Nr.
Makams
# of
Pieces
Usuls
# of
Pieces
Forms
# of
Pieces
1
Hicaz
118
Sofyan
251
Şarkı
677
2
Rast
88
Aksak
246
Türkü
285
3
Nihavent
85
Düyek
143
Seyir
169
4
Uşşak
85
Aksaksemai
101
Küpe
120
5
Segah
74
Curcuna
91
Peşrev
74
6
Hüseyni
72
Ağıraksak
83
Aranağme
72
7
Hüzzam
65
Yürüksemai
75
Sazsemaisi
66
8
Mahur
54
Nimsofyan
74
İlahi
32
9
Kürdilihicazkar
51
Semai
69
Yürüksemai
27
10
Muhayyer
51
Senginsemai
54
Beste
23
Table 2. The most used 10 makams, usuls, and forms in SymbTr
note with the frequency of about 262 Hz and numbered as
60 in the MIDI standard. All notes excluding C’s have a
fractional MIDI Nr. The MIDI Nr corresponding to
CommaAE can be computed by the following formula:
 
 (1)
In order to represent flats and sharps in Hc units,
the "b" and "#" prefixes were used respectively. For ex-
ample, the segah note in AE tone-system is represented
as B4b1; since, according to AE theory, it should sound
one comma lower than the natural B (Si - buselik). Its
comma equivalent is 313, and MIDI Nr is 70.87.
Note53 / Comma53: Indicates the code and the value
of the note in 53TET. If there is no difference between
the performance and the sheet music, CommaAE and
Comma53 values are the same. However, in some makam
sequences such as Uşşak, Hüzzam, Saba and Karcığar,
these two values often vary. For example, in some
makams the pitch that corresponds to B4b1 in AE is
Si4b2 in 53TET, since, in practice, this note should sound
2 commas lower then Si (B). Its comma equivalent is 312,
and MIDI Nr is 70.64.
Numerator / Denominator and ms: Stands for the
rhythmic value of the note, with its duration measured in
milliseconds. When the tempo (quarter note beats per mi-
nute) of the piece is known, these two values can be con-
verted to each other by the following formula:
 
 
 (2)
In Turkish makam music, changes in the tempo of a
piece is a run-of-the-mill situation (e.g., the 4th section of
sazsemaisi pieces are performed faster than other sec-
tions), and since the database can be used for rhythmic
analysis purposes [9], it was found useful to enter these
two strands of information in the same record.
LNS (Legato / Normal / Staccato): Indicates how tied
or detached the notes are to be played. This information is
extracted by listening to performances in synch with
verses and syllables in the lyrics. The default value is 95;
that is, the last 5% of the duration time for normal notes
is completed with silence. 50 means playback should be
of staccato. Rest signs are determined using this value.
VelOn: Indicates the volume or strike of the note,
making nuanced performance possible. Turkish makam
music scores ordinarily do not contain dynamics mark-
ings like piano or crescendo. In SymbTr an attempt has
been made to compensate,
as much as possible, for this
deficiency.
Syllable1: Indicates the
syllable corresponding to a
note. There is one space
character at the end of the
syllables that occur at word
endings and two space char-
acters at the end of the vers-
es. This information was
added to facilitate the track-
ing of the melody, as well as
for its utility in studies of lyrics-based analyses [8]. In
instrumental pieces, it is used to represent the beginning
of sections such as "TESLİM", etc In other places this
field is left blank. Instrumental parts of vocal pieces con-
tain a series of dots in this field. In the original Mus2-
Alpha database, repetitive passages have a separate field
for the second syllable. However, due to copyright con-
siderations there is only one field in SymbTr.
The representation of the score of Fig. 1-c in SymbTr
is listed in Table 1. The data starts immediately after the
column headings. Fields are tab-delimited.
5. MAKAM MUSIC AND MIDI
It is impossible to produce makam music intonations us-
ing ordinary MIDI messages. Therefore, it becomes nec-
essary to use pitch-bend techniques. To generate the
needed feature, a pitch-bend message must be sent with
the same delta-time value as the note, just before the
Note on’ message. The pseudo-MIDI messages for the
first 5 notes in Fig. 1-c are as follows:
Delta
Time
Pitch
Bend
Note
On
0
7 960
C4
4
9 429
D5b
2
9 429
D5b
1
7 960
C4
1
6 492
B
The anchor note is A (La). Therefore, pitch-bend is
unnecessary for any A in all octaves. Bend is required for
all other pitches. For example, the A C interval is 13 Hc
wide. This value is up to 5.7 cents narrower than the
12TET minor third. Taking into account that 100 cents =
4096 pitch bend units, bending for C is calculated as fol-
lows: 8 192 5.7 ∙ 40.96 ≈ 7 960.
MIDI files in SymbTr database are not for listening to
music. They are included, so that the researchers may
find it useful to hear the tune in its simplest raw form. To
this end, even the instrument information has not been
added. Voicing is done with the default MIDI instrument.
6. SOME STATISTICS
SymbTr has been created mainly for the purpose of edu-
cation and scientific research, and hence, endeavored to
be as rich as possible in the diversity of makams, forms,
usuls, and so on. There are many examples such as seyir
composed for educational purposes. One criterion in the
selection of pieces has been music lovers’ familiarity
with them, as to whether a piece be average or above-
average. We did not adopt random sampling (as in [2],
[11], and [15]) as proper methodology when one consid-
ers 80% of the twenty five thousand pieces in the TRT
repertoire have hardly ever been performed or have be-
come obsolete. A musical piece, composed but almost
never performed cannot be held equivalent to one widely
known and frequently performed.
Some statistics about SymbTr as follows:
Total number of pieces: 1 700
Number of notes: ~ 630 000
Classical: 1 400
Folk: 300
Vocal pieces: 1 295
Instrumental pieces: 405
Religious: 49
The number of distinct makams: 155
The number of distinct usuls: 100
The number of distinct forms: 48.
Highest ranking 10 makams, usuls, and forms are
shown in Table 2.
7. PITCHES AND INTERVALS
Of all the pitches in the database, 17 that are used
over 1 per cent in quantity and duration are listed in de-
scending order in Table 3. Percentages in quantity and
duration exhibit slight variances but these do not affect
the ranking.
Figure 4 shows a histogram of these pitches in the
two octave range between yegah (D4: 274) - tizneva (D6:
380) using the note codes as given in Table 3.
It is interesting to note that 9 pitches in the 3 octave
range (Fig. 3) have never been used (for example, kaba-
hicaz, and kabadikhicaz). When we excluded the notes
that were heard for less than one thousandth of the time,
only 33 pitches remained, whereas there were 72 pitches
defined in this range in the AE tone-system. These obser-
vations seem to support Can's results [6].
The most commonly used 13 AE intervals and their
usage as quantity in percentages are listed in Table 4.
The SymbTr database can be accessed at the follow-
ing address, open for public consumption:
http://compmusic.upf.edu
Figure 4. Usage of the notes in SymbTr as durations in percentages
Nr
AE Name
AE Code
Quantity %
Duration %
1
Neva
D5
16.1%
16.1%
2
Çargah
C5
11.0%
10.7%
3
Hüseyni
E5
9.4%
9.7%
4
Gerdaniye
G5
8.5%
9.1%
5
Dügah
A4
8.5%
7.9%
6
Segah
B4b1
6.9%
6.6%
7
Acem
F5
5.4%
5.6%
8
Muhayyer
A5
4.8%
5.3%
9
Eviç
F5#4
4.7%
5.0%
10
Rast
G4
4.0%
3.6%
11
Nimhicaz
C4#4
2.9%
2.8%
12
Hisar
E5b4
1.9%
1.9%
13
Kürdi
B4b5
1.8%
1.7%
14
Dikkürdi
B4b4
1.6%
1.5%
15
Buselik
B4
1.4%
1.4%
16
Nimhisar
D5#4
1.2%
1.2%
17
Dikhisar
E5b1
1.0%
1.0%
Table 3. The most commonly used 17 pitches
Interval
(Hc)
Name, Direction
%
-9
Whole Tone (Tanini), descending
18.1
0
Unison
15.4
-5
Apotome (Küçük Mücennep), desc.
12.5
9
Whole Tone (Tanini)
11.4
5
Apotome (Küçük Mücennep)
8.6
-4
Limma (Bakıyye), desc.
6.3
-8
Minor Whole Tone (B. Mücennep),
desc.
4.2
4
Limma (Bakıyye)
4.1
8
Minor Whole Tone (Büyük Mücennep)
2.6
13
Augmented Second
2.1
-12
Augmented Second, desc.
2.0
-13
Augmented Second, desc.
1.8
22
Perfect Fourth
1.6
Table 4. The most commonly used 13 AE intervals
8. SIMILAR DATASETS
In this article, we announce the availability of a new da-
tabase called SymbTr, the most extensive machine reada-
ble database for Turkish makam music currently availa-
ble. There is only one other compilation that would quali-
fy to be called a database: the recently launched TSM
Corpus [2] (TÜBİTAK
1
ref. is PN: 110K040) consisting
of symbolic data that relate to 600 pieces. These two da-
tabases are far from adequately representing Turkish
makam music. New data, however, is being continually
added to the SymbTr database through various projects. In
addition, Mus2 (Turkish makam and microtonal music
notation program)
2
, which is still being marketed com-
mercially, can produce output in the SymbTr format.
TSM Corpus project, supported by TÜBITAK, can be
quite useful. However, the following deficiencies in data-
base design need to be resolved:
Presence of data belonging to various pieces in a
single Excel format file makes usage difficult,
Syllabized lyrics are not included in the database,
Tempo information for musical pieces is not provid-
ed. Only one quantization information is included
concerning durations: 1/4 meter note = 100 units.
This is a serious drawback for musical pieces that
require, in particular, the inclusion of tempo and / or
usul modulations throughout,
It is not specified which engraved score variant is
employed when entering symbolic data.
9. DISCUSSION
If MIR community members at large run their applica-
tions on the SymbTr database, making necessary small
changes, it may lead to two-way improvements: Myster-
ies of makam music may be unraveled on a grand scale,
at a global setting while scholars keep tapping into new
structures and patterns, thus moving into uncharted terri-
tories of human cognition.
10. ACKNOWLEDGEMENT
This research was partly funded by the European Re-
search Council under the European Union's Seventh
Framework Program, as part of the CompMusic project
(ERC grant agreement 267583).
11. REFERENCES
[1] C. Akkoç, “Non-deterministic scales used in
traditional Turkish music”, Journal of New Music
Research, vol. 31, no. 4, 2002.
[2] N.B. Atalay, S. Yöre: Türk Sanat Müziği derlemi [Data
File]. Retrieved from www.tsmderlemi.com, Konya,
2011.
[3] G. Ay, L.B. Akkal: İTÜ Türk Musikisi Devlet
Konservatuarı Türk müziğinde uygulama - Kuram
1
The Scientific and Technological Research Council of Turkey
2
www.mus2.com.tr
sorunları ve çözümleri - Uluslararası çağrılı kongre
bildiriler kitabı, İstanbul, 2009.
[4] R. Ayangil: Türk makam müziği perdelerini çalabilen piyano
imali projesi. Yıldız Teknik Üniversitesi, İstanbul, 1998.
[5] B. Bozkurt, O. Yarman, M.K. Karaosmanoğlu, C.
Akkoç: Weighing Diverse Theoretical Models on
Turkish Maqam Music Against Pitch Measurements: A
Comparison of Peaks Automatically Derived from
Frequency Histograms with Proposed Scale Tones”,
Journal of New Music Research, 2009.
[6] M.C. Can: “Geleneksel Türk Sanat Müziğinde Arel-
Ezgi-Uzdilek Ses Sistemi ve Uygulamada
Kullanılmayan Bazı Perdeler”, Gazi Üniversitesi, Gazi
Eğitim Fakültesi Dergisi, Ankara, 2001.
[7] M.C. Can: Müzikte bilgisayar destekli istatistiksel analiz,
Ankara, 2004.
[8] H.T. Cheng et al. “Multimodal Structure Segmentation
and Analysis of Music Using Audio and Textual
Information”. EEE International Symposium, 2009.
[9] A. Holzapfel, Y. Stylianou: “Rhytmic Similarity in
Traditional Turkish Music”. ISMIR 2010
<http://www.csd.uoc.gr/~hannover/MMILab-
Andre_files/HolzapfelIsmir09.pdf>.
[10] Y.F. Kutluğ: Türk musikisinde makamlar. YKY, İstanbul,
2000.
[11] N.O. Levendoğlu: “Türk Müziğinde Rast Makamının
İstatistiksel Değerlendirmesi”. Selçuk Üniversitesi Fen
Bilimleri Enstitüsü basılmamış yüksek lisans tezi, Konya,
1996.
[12] G. Oransay: “Das Tonsystem Der Türkei - Türkischen
Kunstmusik”, Musicforschung, Germany, 1959.
[13] İ.H. Özkan: Türk musikisi nazariyatı ve usulleri - Kudüm
velveleleri. Ötüken Neşriyat, İstanbul, 1998.
[14] K.L. Signell: Makam Modal practice in Turkish art music.
Da Capo Press, 1986.
[15] H.T. Sümbüllü, A. Albuz: “Türk Sanat Müziği
dizilerinin, bilgisayar destekli makamsal analizi”,
Uluslararası İnsan Bilimleri Dergisi, cilt 8, sayı 1, 2011.
[16] A. Töre ve E. Karadeniz: Türk musikisi nazariye ve
esasları, T. İş Bankası yayınları, İstanbul, 1965.
[17] Y. Tura: Türk musikisinin meseleleri. Pan Yayıncılık,
İstanbul, 1988.
[18] E. Ünal, B. Bozkurt, M.K. Karaosmanoğlu, N-gram
based Statistical Makam Detection on Makam Music
in Turkey using Symbolic Data”, submitted to ISMIR,
2012.
[19] O. Yarman: "79-tone Tuning & Theory for Turkish
Maqam Music", PhD Thesis, Istanbul Technical University,
Social Sciences Inst., İstanbul, 2007.
[20] R. W. Young: "Terminology for Logarithmic
Frequency Units". The Journal of the Acoustical Society of
America 11 (1), 1939.
... We simulate this data by comparing the implicit rhythm categorization of two pPIPPET filters, and relate the results to empirical observations. Alongside the filter created in Experiment 2 using monophonic German folksongs [46], we created a pPIPPET filter with expectations derived from monophonic Turkish makam music [48]. We refer to these filters as the German and Turkish models, but again stress that the music corpora used to train these models only approximate the listening experience of individuals from these musical cultures. ...
... We have presented pPIPPET, a model of entrainment to a time series of events, which draws upon prototypical patterns of temporal expectations in order to accurately track the event Results from the final iteration of all simulated trials, using pPIPPET filters configured with either German folk songs [46] or Turkish makam music [48]. A) Kernel density estimate (KDE) of the underlying data distribution for the German model, using the non-parametric method described in [24], normalized relative to a uniform distribution. ...
... Metrical models. We leverage the metrical analysis by Van der Weij [15], which compares the statistical properties of monophonic score-based rhythms in German folk melodies [46] and Turkish makam music [48]. The rhythm samples analyzed were carefully curated to ensure an equal number of total rhythms, which had been truncated to segments of uniform length, and filtered to only include rhythms defined at a sixteenth-note resolution. ...
Article
Full-text available
Long-term and culture-specific experience of music shapes rhythm perception, leading to enculturated expectations that make certain rhythms easier to track and more conducive to synchronized movement. However, the influence of enculturated bias on the moment-to-moment dynamics of rhythm tracking is not well understood. Recent modeling work has formulated entrainment to rhythms as a formal inference problem, where phase is continuously estimated based on precise event times and their correspondence to timing expectations: PIPPET (Phase Inference from Point Process Event Timing). Here we propose that the problem of optimally tracking a rhythm also requires an ongoing process of inferring which pattern of event timing expectations is most suitable to predict a stimulus rhythm. We formalize this insight as an extension of PIPPET called pPIPPET (PIPPET with pattern inference). The variational solution to this problem introduces terms representing the likelihood that a stimulus is based on a particular member of a set of event timing patterns, which we initialize according to culturally-learned prior expectations of a listener. We evaluate pPIPPET in three experiments. First, we demonstrate that pPIPPET can qualitatively reproduce enculturated bias observed in human tapping data for simple two-interval rhythms. Second, we simulate categorization of a continuous three-interval rhythm space by Western-trained musicians through derivation of a comprehensive set of priors for pPIPPET from metrical patterns in a sample of Western rhythms. Third, we simulate iterated reproduction of three-interval rhythms, and show that models configured with notated rhythms from different cultures exhibit both universal and enculturated biases as observed experimentally in listeners from those cultures. These results suggest the influence of enculturated timing expectations on human perceptual and motor entrainment can be understood as approximating optimal inference about the rhythmic stimulus, with respect to prototypical patterns in an empirical sample of rhythms that represent the music-cultural environment of the listener.
... The fundamental features of music of the West started to appear in classical Turkish music, as well, with the start of transcribing notes. One example is the manner in which Rast makam sequence was played on piano with the notes closely following the C-Major scale of the West (Karaosmanoğlu, 2012;Yarman, 2008). However, the major or minor scales of the West still do not have exact counterparts in classical Turkish music, which is based on quite different structures known as "usül" (style) and makam (Gedik and Bozkurt, 2010;Akkoç, 2002). ...
... In Western Music literature, Major notes were associated with happier emotions, while Minor notes were associated with sad emotions (Davidson, Scherer ve Goldsmith, 2002). The emotions invoked by Rast melodies show parallels with the emotions raised by the major scales of Western music (Karaosmanoğlu, 2012;Yarman, 2008), while Hüzzam melodies had emotional invocations similar to those in the minor scales of the Western music. But yet even in Hüzzam and in Rast Makam faster rhythms were found to be happier. ...
... Traditional Turkish Music, which is classified under two major headings, as Turkish Folk Music and Classical Turkish Music, has modal and monodical features. According to Karaosmanoğlu (2012), it is a genre drawing roots from a thousand year old tradition, featuring distinct melodic patterns called makam and rich rhythmic structures called usul. Yekta (1924) defines makam as a specific form of a musical scale that characterizes itself by an organization of intervals and various constitutive relations (as cited in Bozkurt et al., 2014). ...
... Perde adları olarak Arel ‐ Ezgi ‐ Uzdilek sistemin‐ dekiler kullanılmıştır ve bir dosya içinde 5 eserin verileri bulunmaktadır. Bu tezin yazarı tarafından hazırlanıp paylaşıma açılan ikinci derlem SymbTr (Karaosmanoğlu [93]) anons edildiğinde 155 makamdan 1700 eser içeriyordu. Koleksiyona adını veren SymbTr, gerçekte Türk musikisi eserlerini sembolik olarak eksiksiz biçimde temsil etmek üzere tasarlanmış bir metin formatıdır. ...
Thesis
Full-text available
COMPUTATIONAL MELODIC ANALYSIS ON SYMBOLIC DATA OF TURKISH MAKAM MUSIC Mustafa Kemal KARAOSMANOĞLU Department of Mathematics Engineering Phd. Thesis Adviser: Prof. Dr. Fatih Taşçı Co‐Adviser: Assoc. Prof. Dr. Barış Bozkurt The discipline of computational musicology which can be summarized as the modelling and simulation of music with the methods of mathematics has been used to analyse the music of many cultures. The processes that would take long time or sometimes even impossible to do manually can easily be accomplished by using computer support using these techniques. These techniques which offer useful tools in areas such as having a better understanding of music, its learning and teaching, its performing, its comparison with other kinds of music, its access on the Internet have rarely been studied for Turkish music. With this study we have tried to enrich this field which is untouched for reasons such as the unique character of this music, the fact that it had not been modelled and thus had not been entrenched with the appropriate data. The text of the thesis starts with chapter describing the relational database structure created by the author which enables, through the system analysis techniques, the structuring, modelling and simulation of Turkish makam music. This stage has to do with construction of the symbolic database by which the computational musicology techniques will be applied to Turkish makam music. As with the application on these data i) recognition of the makam with n‐gram technique ii) The detection of specific makam tunes with machine learning techniques iii) improving automatic melodic segmentation algorithms are studied in depth. A number of standards such as MIDI for music that use 12 tone equal temperamant (12TET) system sounds have been defined long ago and researchers have tried using computational techniques on data which are within these standards. But for the Turkish makam music, it is not possible to start working directly using these standards. Because the notes / pitches that can be called the building blocks of this music are very different from those with 12TET both as numerically and as of value as well as data structures. These differences can be traced looking at the diversity of theoretical propositions that have been called to attention especially from the beginning of the 20th century. In the thesis first of all the system and the format to represent the Turkish music suggested by author is described and extensive database in this format has been introduced. As an exemplary application for the use of techniques computational musicology makam recognition, with n‐gram technique and the patterns of specific makam tunes with statistical decision theory method have been described and especially automatic melodic segmentation with machine learning techniques have deeply been treated. It has been shown that the results obtained through the use of structures specific to Turkish music are superior to the algorithms in the literature. Keywords: music, computational musicology, Turkish music, segmentation, melodic analysis, musical intervals, sound system
... i) Composed in the most commonly used makams (Çevikoğlu, 2007): Acemaşiran, Beyati, Buselik, Hicaz, Hicazkar, Hüseyni, Hüzzam, Kürdilihicazkar, Mahur, Muhayyer, Neva, Nihavent, Rast, Saba, Segah and Uşşak, Overall, a set of 480 pieces was collected consisting of 30 pieces for each of the 16 distinct makams by rewriting the pieces using Mus2 microtonal notation software (http://www.mus2.com.tr/) in the Arel notation (Arel, 1968) and further converting this data to the machine readable text format of SymbTr (Karaosmanoğlu, 2012). Three experts were asked to mark the phrase boundaries and çeşni/geçki modulations on printed scores, as they would do it for makam melodic analysis. ...
Conference Paper
Full-text available
1. ABSTRACT One of the basic needs for computational studies of traditional music is the availability of free datasets. This study presents a large machine-readable dataset of Turkish makam music scores segmented into phrases by experts of this music. The segmentation facilitates computational research on melodic similarity between phrases, and relation between melodic phrasing and meter, rarely studied topics due to unavailability of data resources.
Article
Experienced listeners internalize musical tonal knowledge via statistical learning of pitch distributions as a result of exposure to musical environment. Cross-cultural studies of music cognition offer new perspectives to investigate the acquisition of tonal schema. Makam music is a rich musical system characterized by modal structures defined by micro-tonal pitch sets, and melodic progression patterns (aka seyir features). Makam schema is possibly acquired by internalizing the seyir in addition to pitch features. In the current study, we examined whether an ideal model of makam schema is built with multidimensional scaling analysis and with self-organizing maps (SOMs). We were interested in whether statistical information about seyir features, in addition to pitch distributions, would form an acceptable makam schema model. We qualitatively analyzed topographical organizations in the models to understand whether they reflect complex relations between makams. Multidimensional scaling analyses did not produce an acceptable model for makam schema. The SOM trained with pitch distributions provided an adequate model for makam schema. However, the SOM trained with both pitch distributions and seyir features was better in capturing the complex relations between makams. Further behavioral research is necessary to understand whether melodic progression patterns are intrinsic features of the tonal knowledge of the experienced listeners of makam music.
Research
Full-text available
This project is dedicated to development of automatic melodic segmentation and analysis algorithms and tools for Turkish makam music and perform detailed testing of the developed tools. These efforts have led to two novel algorithms; one for automatic melodic segmentation and one for automatic classification of makam melodies. For data collection, 1000 pieces, selected to have equal distribution over the historical periods and most frequently used makams, were written using a microtonal notation software, then converted to various formats for analysis. 500 of these pieces were then manually segmented into phrases by three experts of makam music. The first target was to develop an automatic algorithm using this data, available algorithms, additional features specific to makam music and machine learning tools for melodic segmentation. The developed algorithm was tested using the standard F-measure and the results confirmed higher efficiency for the proposed algorithm compared to algorithms from the literature. This algorithm was then used to automatically segment the resting 500 pieces. The project has reached all its goals with additional outputs in the domain of seyir analysis and makam analysis. All collected data and developed tools are shared on the web-site: http://akademik.bahcesehir.edu.tr/~bbozkurt/112E162.html or alternatively https://dl.dropboxusercontent.com/u/46516299/112E162.html Keywords: Melodic analysis, automatic melodic segmentation, makam music, Turkish music
Article
Full-text available
Musical information retrieval (MIR) applications have become an interesting topic both for researchers and commercial applications. The majority of the current knowledge on MIR is based on Western music. However, traditional genres, such as Classical Turkish Music (CTM), have great structural differences compared with Western music. Then, the validity of the current knowledge on this subject must be checked on such genres. Through this work, a MIR application that simulates the human music processing system based on CTM is proposed. To achieve this goal, first mel-frequency cepstral coefficients (MFCCs) and delta-MFCCs, which are the most frequent features used in audio applications, were used as features. In the last few years deep belief networks (DBNs) have become promising classifiers for sound classification problems. To confirm this statement, the classification accuracies of four probability theory-based neural networks, namely radial basis function networks, generalized regression neural networks, probabilistic neural networks, and support vector machines, were compared to the DBN. Our results show that the DBN outperforms the others.
Article
Full-text available
This work studies the effect of different score representa-tions and the potential of n-grams in makam classification for traditional makam music in Turkey. While makams are defined with various characteristics including a dis-tinct set of pitches, pitch hierarchy, melodic direction, typical phrases and typical makam transitions, such cha-racteristics result in certain n-gram distributions which can be used for makam detection effectively. 13 popular makams, some of which are very similar to each other, are used in this study. Using the leave-one-out strategy, makam models are created statistically and tested against the left out music piece. Tests indicate that n-gram based statistical modeling and perplexity based similarity metric can be effectively used for makam detection. However the main dimension that cannot be captured is the overall progression which is the most unique feature for classifi-cation of close makams that uses the same scale notes as well as the same tonic.
Article
Full-text available
Non-deterministic pitch scales observed in traditional Turkish music are viewed as distributions along the pitch axis. This thesis contrasts with the conventional norm of using deterministic scales, where pitches are discrete fixed points on the number line (the pitch axis), as in fretted and keyboard instruments. In this study actual frequencies used by master musicians during an improvisation are measured in Hertz and analyzed with the long term goal of characterizing the underlying modal scales in the form of distributions. The potential of the measurement system, together with a variety of plausible mathematical analysis schemes is demonstrated on improvisations made by two prominent master musicians in a specific fundamental mode.
Conference Paper
Full-text available
In this paper, we present a multimodal approach to structure segmentation of music with applications to audio content analysis and music information retrieval. In particular, since lyrics contain rich information about the semantic structure of a song, our approach incorporates lyrics to overcome the existing difficulties associated with large acoustic variation in music. We further design a constrained clustering algorithm for music segmentation and evaluate its performance on commercial recordings. Experimental results show that our method can effectively detect the boundaries and the types of semantic structure of music segments.
Article
Since the early 20th century, various theories have been advanced in order to mathematically explain and notate modes of Traditional Turkish music known as maqams. In this article, maqam scales according to various theoretical models based on different tunings are compared with pitch measurements obtained from select recordings of master Turkish performers in order to study their level of match with analysed data. Chosen recordings are subjected to a fully computerized sequence of signal processing algorithms for the automatic determination of the set of relative pitches for each maqam scale: f0 estimation, histogram computation, tonic detection þ histogram alignment, and peak picking. For nine well-recognized maqams, automatically derived relative pitches are compared with scale tones defined by theoretical models using quantitative distance measures. We analyse and interpret histogram peaks based on these measures to find the theoretical models most conforming with all the recordings, and hence, with the quotidian performance trends influenced by them.
Conference Paper
In this paper, the problem of automatically assigning a piece of traditional Turkish music into a class of rhythm referred to as usul is addressed. For this, an approach for rhyth- mic similarity measurement based on scale transforms has been evaluated on a set of MIDI data. Because this task is related to time signature estimation, the accuracy of the proposed method is evaluated and compared with a state of the art time signature estimation approach. The results indicate that the proposed method can be successfully ap- plied to audio signals of Turkish music and that it captures relevant properties of the individual usul.
Türk Sanat Müziği derlemi [Data File
  • N B Atalay
  • S Yöre
N.B. Atalay, S. Yöre: Türk Sanat Müziği derlemi [Data File]. Retrieved from www.tsmderlemi.com, Konya, 2011.
79-tone Tuning & Theory for Turkish Maqam Music
  • O Yarman
O. Yarman: "79-tone Tuning & Theory for Turkish Maqam Music", PhD Thesis, Istanbul Technical University, Social Sciences Inst., İstanbul, 2007.
Akkal: İTÜ Türk Musikisi Devlet Konservatuarı Türk müziğinde uygulama -Kuram [4] R. Ayangil: Türk makam müziği perdelerini çalabilen piyano imali projesi
  • G Ay
G. Ay, L.B. Akkal: İTÜ Türk Musikisi Devlet Konservatuarı Türk müziğinde uygulama -Kuram [4] R. Ayangil: Türk makam müziği perdelerini çalabilen piyano imali projesi. Yıldız Teknik Üniversitesi, İstanbul, 1998.