Algomus datasets

http://www.algomus.fr/data

This repository contains and links public data released by the Algomus team. Open data and corpora are essential for digital humanities studies. Systematic musicology approaches use corpora to infer new music knowledge or to confirm or challenge existing theories. Computational music analysis (CMA), or, more generally, music information retrieval (MIR) studies design algorithms that need to be evaluated.

Building corpora with reference analyses is not so easy – there are many different analyses of the same piece that are musically relevant. However, some analysis are definitely more correct than others. Some reference annotations can be used to evaluate some parts of analysis algorithms. As computer or data scientists, we would like to have computer-readable reference datasets that may be used as a ground truth to evaluate MIR/CMA algorithms or to infer new music knowledge. But as music theorists, we know that there is not only one correct analysis of a given piece: listeners, players, or theorists often disagree or at least propose several points of view. Anyway, there is consensus about some analytical elements by many music theorists, players or listeners. The fact that reaching consensus may be difficult on some points should not prevent us from trying to formalize some elements.

Available datasets

Fugues

Data: https://gitlab.com/algomus.fr/algomus-data/tree/master/fugues

24 Bach fugues (WTC I, BWV 846-893) + 12 Shostakovich fugues (op.57, 1952)
S/CS/CS2 patterns, cadences, pedals (1000+ labels)
.ref (2013 – 2016) + converted .dez (2019)

Reference: M. Giraud et al., Computational Fugue Analysis, Computer Music Journal, 39(2), 77-96, doi:10.1162/COMJ_a_00300, 2015, https://hal.archives-ouvertes.fr/hal-01113520

Texture in string quartets

Data: https://gitlab.com/algomus.fr/algomus-data/tree/master/quartets/texture

11 movements (Haydn, Mozart, Schubert)
700+ texture labels
.ref (2014 and 2021), also translated from .dez files (2021)

Texture in piano music

Data: https://gitlab.com/algomus.fr/symbolic-texture-dataset

9 movements from Mozart piano sonatas (K279, K280 and K283)
1164 texture labels
.dez, .txt (human-readable format), .tsv (machine-readable format)

References:

L. Couturier et al., A Dataset of Symbolic Texture Annotations in Mozart Piano Sonatas, ISMIR 2022
L. Couturier et al., Annotating Symbolic Texture in Piano Music: a Formal Syntax, SMC 2022

Mozart String Quartets

Data: https://gitlab.com/algomus.fr/algomus-data/tree/master/quartets/mozart

32 sonata-form and sonata-form-like movements
sonata form structure (following Hepokoski&Darcy 2006) and cadences (2000+ labels)
.dez encoded within Dezrann by L. Feisthauer (2017 – 2019)

Additional data: features occurrences, models with learned probabilities (2019)

Reference: P. Allegraud et al., Learning sonata form structure on Mozart’s string quartets, in revision

Orchestration

Data: https://gitlab.com/algomus.fr/orchestration

24 first movements of Haydn, Mozart, and Beethoven symphonies
7941 texture labels describing 8528 bars

Reference: D. V. T. Le et al., A Corpus Describing Orchestral Texture in First Movements of Classical and Early-Romantic Symphonies, DLfM 2022

Guitar positions

Data: https://gitlab.com/algomus.fr/algomus-data/tree/master/guitar/

2248 tracks in 1022 songs/pieces
1000 most frequent position vectors for chords (with 2-3 notes, with 4-6 notes)

Reference: J. Cournut et al., What are the most used guitar positions?, DLfM 2021

Data: https://gitlab.com/lbigo/rhythm-guitar-detection

For 102 guitar tablatures, at the bar level:

Manual annotations (rhythm guitar or not)
31 computed features (note/chord values, variety, playing techniques, etc.)

Reference: D. Regnier et al., Identification of rhythm guitar sections in symbolic tablatures, ISMIR 2021

Data formats

Data are presented in several formats, including:

.dez json files, that can be read and written
- by Dezrann
- by our music21.schema module (see Bagan 2015, upcoming update in 2019)
.ref annotation files, described in syntax.ref

We publish all new data in .dez format, and will progressively convert the old files.

License

These data are made available under the Open Database License. Any rights in individual contents of the database are licensed under the Database Contents License.

Other data

AI Song Contest

https://gitlab.com/algomus.fr/algomus-data/tree/master/ai-song/2020-i-keep-counting

Reference: G. Micchi et al., I Keep Counting: An Experiment in Human/AI Co-creative Songwriting, TISMIR, in press