We are always looking for students to collaborate with, routinely supervising Master’s theses at the intersection of musicology and computing. We also offer part-time employment opportunities (student assistantships), as well as for-credit semester projects.
The following projects are currently proposed by various lab members. If you find any of them appealing, please contact Dr. Yannis Rammos or the designated researcher, laying out your interests as well as relevant background. You are also welcome to submit your own research proposal to us, initially in the form of a few sentences.
Conducting behavioral experiments
Semester project Student assistantship
A number of behavioral experiments are already designed and await execution: participants need to be recruited and brought to the lab, experiments performed, and data recorded.
Contact: Gabriele Cecchetti
Extending our music processing library
Semester project
We are in the process of building up an extensive music processing toolkit by integrating existing frameworks as well as developing new functionality. (Our dimcat library, for example, is part of this effort.) Apart from basic audio processing functionality, the toolkit’s focus is on audio-to-symbolic transcription of music and music processing and analysis on the symbolic level. This includes things like:
- beat inference
- measure inference
- harmonic inference
- motive analysis
- integrating or reproducing the state-of-the-art, among others
- matlab midi toolbox
- music21 python framework
- Essentia
- reading/writing/converting between various commonly used formats
- “rapid prototyping” (integration with visualization and sonification functionality from above)
Contact: Johannes Hentschel
Investigating a perceptual reality behind music-theoretical structure
Master's thesis
A central aspect of music theory is identifying and analyzing structures in music. Some of these structures can be described with syntactical models akin to those used in language. Whether or not these theoretical models allow prediction on a perceptual level, or are predominantly useful on an analytical (interpretive) level is the subject of the present project. In this project, you will design, implement, conduct, and analyze an empirical experiment that aims to assess the viability of music-theoretical claims as perceptual hypotheses. In most cases the tested hypotheses will be related to harmony, rhythm, long-distance relationships between musical events, or perception of closure.
Prerequisites: Fundamental knowledge in data science / statistical data analysis, skill in an appropriate programming language (e.g., R, Python, or MatLab), interest in experimental psychology and music.
Contact: Gabriele Cecchetti
“Hidden motive” discovery (subgraph mining)
Master's thesis
“Melodic motives” are successions of notes which recur in a musical piece, thus contributing to its individual character. Like the familiar four notes opening Beethoven’s “Fifth”, motives are generally easy to identify, both aurally and computationally.
Also recognized, albeit less researched, is a class of “hidden motives” which consist of non-consecutive notes (subject to certain harmonic constraints). Despite their shadowy nature, hidden motives are considered to be an essential feature of Western classical-music aesthetics, and a means of coherence, psychological complexity, and dramatic conviction. It takes meticulous study for a music analyst to discover hidden motives within a piece—the process is akin to a musical “treasure hunt.” A computational method for their discovery would be valuable in musicology, artistic practice, music education, and music cognition research.
From a computer science point of view, this challenge is a nontrivial “Frequent Subgraph Mining” (FSM) problem: both the (initially unknown) motives, and the musical work within which they hide, are represented as mathematical graphs encoded in an XML-compliant syntax. Heuristic, combinatorial-optimization, and deep-learning approaches to FSM may be considered as points of entry into the mathematical aspects of the problem.
Prerequisites:
- mathematical graph theory
- music reading skills
- Python programming
- some knowledge of Western-music harmony and voice-leading (desirable)
Indicative bibliography:
- Aziz, A. (2018). Beyond ‘Three Blind Mice’: An Exemplar of ‘Day 1’ of Schenkerian Analysis. Journal of Music Theory Pedagogy (online resources).
- Cadwallader, A., Gagné, D., Samarotto, F. (2019). Analysis of Tonal Music: A Schenkerian Approach. Oxford University Press.
- Dhiman, A., & Jain, S. (2016). Optimizing Frequent Subgraph Mining for Single Large Graph. Procedia Computer Science, 89, 378-385.
Contact: Yannis Rammos
Modeling the note level of jazz harmony
Master's thesis
Harmonic analyses in the form of chord annotations represent a coarse-grained segmentation of the score with a given chord label expressing an abstraction of a particular segment. Such abstract chord labels therefore enable researchers to investigate differences and common patterns in different tonal languages. The DCML has access to a small corpus of Jazz transcriptions with included chord labels (changes). The aim of this thesis would be to relate the abstract labels to the concrete notes that the performers played in the respective segments, and to look for significant stylistic differences in various performers’ realization of particular chords.
The project is aimed at a highly motivated student interested in jazz harmony and its realization in particular jazz recordings. It requires music theoretical knowledge and very good skills in programming and data analysis. Prior experience with machine learning is an asset but not a requirement. Methodological choices are to be discussed.
Contact: Johannes Hentschel
Music Explorer: Visualizing chronicles of musical “facts”
Semester project Student assistantship
As part of our lab’s overarching Corpus Initiative and the ongoing Distant Listening project, our lab recently prototyped MusicExplorer, a web app for visualizing large corpora of music analyses and related historiographical information. The app is written in Plotly Dash and effectively serves as a frontend of dimcat, our Digital Musicology Corpus Analysis Toolkit. Building on the existing prototype, we now intend to develop additional user interfaces which should enable users to formulate significantly more complex and insightful queries.
Prerequisites:
- Python mastery (preferably with experience in Plotly Dash);
- dedication to writing elegant, maintainable, well-documented code;
- expertise or strong interest in UI/UX design;
- good technical-writing skills in English.
Contact: Yannis Rammos
Tradition and innovation in the notation of linear analysis
Semester project Master's thesis
Over one century since their inception, linear techniques of music analysis are still among the most revelatory and expressively powerful instruments available for the interpretation of classical music. Linear analyses—based on the techniques of Leo Mazel, Heinrich Schenker, or Célestin Deliège, among several others—capture “deep structures” of a musical work, often using extended score notations which pose challenges to common techniques of symbolic music encoding and score rendering. Our lab has already developed an XML-based representation of such analyses, an interface for analysis encoding, encoding guidelines, as well as a corpus of canonical analyses (soon to be published). This project has a two-pronged goal: first, to develop a Javascript renderer of XML-encoded analyses in score notation, building on the Verovio library; second, to design radically innovative graphic representations of these analyses, departing from those of historical theorists, potentially borrowing ideas from network visualization or even cartography.
Prerequisites:
- interest in digital music typography and creative visualization
- expertise in vector graphics techniques (incl. SVG)
- fluency in mathematical graphs
- fluency in score reading
- basic knowledge of Western tonal harmony and principles of polyphony
- excellent Javascript/ES6 or Python skills
Indicative bibliography:
- Ericson, P., Rammos, Y., & Rohrmeier, M. (in press). A Generic Framework for Hierarchical Music Analysis. Music Encoding Conference Proceedings 2022.
- Gould, E. (2011). Behind Bars: The Definitive Guide to Music Notation. Alfred Music.
- Schenker, H. (1969). Five Graphic Analyses. Dover. (Original work published 1932.)
Contact: Yannis Rammos
Rich Classical-Music Metadata: A Case-Study from Verbier
Semester project
In comparison with other genres, descriptive data about classical music recordings—otherwise known as “metadata”—present special musicological and technological challenges. These are routinely evidenced in discoverability problems of music search engines, music librarianship, and streaming platforms. Performance research, archival research, music historiography, and music AI are also impeded by the shortage of high-quality metadata for classical performance.
In this rather typical Digital Humanities project, we aim at fundamentally rebuilding the metadata database of the Verbier Festival archive, a world-class collection of contemporary, historical, and indeed historic performances, which continues to grow 57 years since the festival’s inauguration. In its current state, the database is an incomplete listing of ad hoc data entries, developed in utilitarian fashion to serve specific needs of the festival without a firm commitment to data design and archiving principles. The project thus necessitates expertise in data ontology design, musicological sensibility, and extensive “data wrangling.” Building on existing open-data resources, our goal is to:
- design a musicologically sound, technologically sustainable schema for classical music performance metadata;
- turn this schema into a “compound” data ontology which will also include, or link to, metadata for printed scores, thus coupling live concert performances with the musical texts actually performed;
- use this ontology to build a metadata database for the Verbier Festival recordings archive (the scope of this database is to be determined, potentially in collaboration with the EPFL Cultural Heritage Center, which acts as our institutional interface with the festival);
- produce a report with descriptive statistics and insights drawn from them.
The third step will likely involve the use of OpenRefine to reconcile already available metadadata with content drawn from “authoritative” open-data sources, such as WikiData.
Prerequisites:
- understanding of data-design principles (ontologies, XML, RDF, OWL, etc.), or strong motivation for self-study in the area;
- experience or interest in structured-data sources (e.g. WikiData) and data wrangling software (e.g. OpenRefine);
- proficiency in regular expressions (RegEx);
- data-analysis skills, including experience in Python Pandas and Plotly.
- meticulousness, eye for detail in large data;
- interest in music historiography or classical music (optional but desirable).
Contact:
Cédric Duchêne, EPFL+ECAL Lab
Delphine Ribes Lemay, EPFL+ECAL Lab
Yannis Rammos, EPFL Digital & Cognitive Musicology Lab