🇫🇷 Français
🇬🇧 English

PhD 2021-2024 – Modeling musical texture for computer-aided analysis and composition

  • Fully funded PhD 2021-2024 in computational music analysis, algorithms, and machine learning
  • Lieu: Amiens (UPJV, Laboratoire MIS), collaboration with UniversitĂ© de Lille (CRIStAL, CNRS)
  • Supervisors and contacts: Florence LevĂ© (MdC HDR MIS, UPJV), Louis Bigo (CRIStAL)
  • Deadline for applications 30 april 2021 (CV and motivation letter, by mail)

Context

Music is made of melodies and harmonies structured in time. The musical score is an essential support for transmitting, sharing and preserving Western music. Today, computational approaches combining musical expertise, algorithmic techniques and machine learning allow to classify, analyze and generate music. These advances make it possible to consider the realization of applications where human and machine collaborate in the analysis or writing of music.

Most existing approaches, based on rules or learning methods, consider the ‘note’ as the main object to model harmony and tonality. On a higher level, music can be characterized by its ‘texture’: solo, accompaniment, homorhythm or imitation are examples of textural processes that the composer uses when writing a piece. The combination of voices (alone, in group) according to these different textures plays an important role in our perception of music. The computational modeling of texture has been the subject of preliminary studies [Giraud2014, LevĂ©2017]. The objective of this thesis is to study and compare the computational modeling of musical texture in different corpora of scores:

  • The classical string quartets of Haydn, Mozart and Beethoven
  • The MySongBook repertoire (guitar) maintained by the company Arobas Music with which the Algomus team collaborates.
  • A corpus of piano music (e.g. Maestro, maintained by the Google Magenta team). This modeling could be used as a basis for the development of co-creativity tools, helping in the composition or pedagogy of musical composition.

Objectives

  • Production of texture reference analyses on several musical repertoires.
  • Development of algorithms allowing the textural analysis of musical pieces of the chosen repertoires.
  • Validation of the proposed algorithms by comparing their results with the reference analysis previously performed.
  • Evaluation of the capacity of deep neural networks to improve the analysis of musical texture
  • Evaluation of the ability of unsupervised learning models, used in particular in the field of pattern detection, to discover the preponderant textural elements in the corpora for which we have texture annotations.
  • Application of texture recognition algorithms on large corpora to build consistent sub-corpora (melody corpora, accompaniment corpora, imitation corpora, etc.) which will be made available to the MIR community for the specific study of these different musical notions and for music generation.
  • Development of textural scenarios and methods to help texture-driven music composition.

The corpora, models and tools created during this thesis will be freely distributed. Emphasis will be put on the creation of reusable tools, namely a module allowing both beginners and advanced musicians to experiment with texture-based analysis and composition. A module for visualization and interaction with the analysis will be made available via the Dezrann platform, allowing musicians as well as the general public to study this texture in a variety of music, as well as to college classes in the context of a pedagogical workshop. For all these objectives, the thesis will be part of a network of collaborations woven by the MIS and the CRIStAL over the last few years, including academic collaborators in computer music as well as in musicology, music teachers, and R&D laboratories.

Profile of the candidate

MSc. in Computer Science, with skills in Machine Learning, Data Science and algorithms. Musical Practice and knowledge in musical theory (analysis, composition…)

Schedule, publications, and collaborations

The team is used to publish at major conferences in the field, including ISMIR. The PhD student will be guided in writing and submitting papers to these conferences, whether on his/her own work or on collaborative work inside or outside the team. The PhD student will regularly participate at conferences and other events in the field, and will be encouraged to collaborate with other scientists and artists. Each PhD student in the team also undertakes an international research stay during the course of his or her thesis.

The schedule is individualized for each student, but could be as follows:

  • T0 to T0+6 months: bibliography
  • T0+6 to T0+16: first models and algorithms, first paper submission
  • During the second year: 2-months stay in a foreign partner university to consolidate the work and open up to new projects
  • T0+18 to T0+26: further model and algorithm design, implementation, and eveluation, submissions on the personal work as well on collaborative work
  • T0+26 to T0+34: thesis writing, possibly new paper submission
  • T0+35 and T0+36: PhD defense, new projects and collaborations

References

  • [Briot2019] J. P. Briot, G. Hadjeres, F. Pachet, Deep Learning Techniques for Music Generation - A Survey, 2019, https://arxiv.org/abs/1709.01620
  • [Bigo2017] L. Bigo et al., Sketching Sonata Form Structure in Selected Classical String Quartets, ISMIR 2017, https://hal.archives-ouvertes.fr/hal-01568703/document
  • [Duane2012] B. Duane, Texture in Eighteenth- and Early Nineteenth- Century String-Quartet Expositions, PhD thesis, North-western University Evanston, 2012
  • [Esling2020] P. Esling, N. Devis, Creativity in the era of artificial intelligence, https://arxiv.org/abs/2008.05959
  • [Giraud2014] M. Giraud, F. LevĂ© et al., Towards Modeling Texture in Symbolic Data, ISMIR 2014, https://hal.archives-ouvertes.fr/hal-01057017/document
  • [Komlos 1986] K. Komlos, Haydn’s Keyboard Trios Hob. XV:5-17: Interaction between Texture and Form, Studia Musicologica Academiae Scientiarum Hungaricae, 28:1-4, 351-400, 1986
  • [LevĂ©2017] F. LevĂ©, M. Rigaudière, F. DoĂ©, Vers une analyse informatique des textures dans le quatuor classique, EuroMAC 2017
  • [Levy1982] J. M. Levy, Texture as a Sign in Classic and Early Romantic Music, Journal of the American Musicological Society, Vol. 35:3, 482-531, 1982
  • [Medeot2018] G. Medeot, et al., StructureNet: Inducing Structure in Generated Melodies, ISMIR 2018, http://ismir2018.ircam.fr/doc/pdfs/126_Paper.pdf
  • [Trimmer1981] M. A. Trimmer, Texture and sonata form in the late string chamber music of Haydn and Mozart, PhD thesis, 1981