C4DM Seminar: Computational Recognition of Emotions in Music: A Meta-Analysis and Critical Review

This talk presents a meta-analysis of music emotion recognition (MER) models published between 2014 and 2024, focusing on predictions of valence, arousal, and categorical emotions. A total of 553 studies were identified, of which 96 full-text articles were assessed. This resulted in a final review of 34 studies comprising 204 distinct models. Valence and arousal were predicted with reasonable accuracy (r = 0.67 and r = 0.81, respectively), while classification models achieved an accuracy of 0.87. Across modeling approaches, linear and tree-based methods generally outperformed neural networks in regression tasks, whereas neural networks and support vector machines showed the highest performance in classification tasks. The talk will address several critical issues, including conceptual shortcomings, insufficient quality control, lack of transparency, limited dataset size, and a narrow range of predicted emotions in computational approaches to music emotion recognition.

Tuomas Eerola is a music psychologist and Professor of Music Cognition at Durham University, UK. He is a leading researcher in the field of music cognition, using empirical experiments, theorising, and computational models to study how people engage with and process music. His work encompasses a wide range of topics, including emotions induced by music and perception of musical structure, rhythm, timbre, consonance, and emotional communication through music. He has published more than 160 papers and book chapters and is on the editorial boards of several prominent music psychology journals. His research has been funded by the Academy of Finland, AHRC (UK), ESRC (UK), and EU Horizon 2020. Eerola is the author of the book "Music and Science - Guide to Empirical Research" published by Routledge in 2024.

University profile: https://www.durham.ac.uk/staff/tuomas-eerola/
GitHub: https://tuomaseerola.github.io
Music & Science Lab: https://musicscience.net