Public defence: Computational Music Performance Analysis: A multimodal, context-informed approach

This PhD project investigates how multimodal datasets can support the computational analysis of music performances. These datasets combine different types of information, including audio, video, movement, text, musical scores, and physiological data. The research focuses on how machine-learning methods can combine complementary data sources to address specific analytical tasks. It also examines practical and ethical issues in building multimodal music datasets, including categorization, evaluation, and cultural and legal considerations. 

The dissertation introduces two datasets. MusiQAl is designed for music question-answering and combines audiovisual information with contextual and cultural knowledge about performances. The Greek Folk Music Dataset (GFMD) combines MIDI transcriptions, aligned lyrics, and metadata to support computational analysis of Aegean dance music. Together, these datasets show how multimodal and context-aware approaches can provide new ways to analyze music performance and establish a methodological basis for further computational research in music.

Anna-Maria Christodoulou successfully defended her thesis on 18 September 2026. 

Trial lecture

Designated topic: Music Understanding: Interdisciplinary synergies in machine and human cognition of music. Time and place: Friday 18 September, 11.15. Forsamlingssalen, Harald Schjelderups hus

Thesis introduction

First opponent

Second opponent

Evaluation committee

• Professor Anja Volk, Utrecht University (first opponent) 
• Professor Xavier Serra, Universitat Pompeu Fabra (second opponent)
• Associate Professor Tejaswinee Kelkar, University of Oslo (committee administrator) 

Chair of the defence

Supervisors