Maryam Fayaz Torshizi
Music mood modelling using Knowledge graphs and Graph Neural Nets.
Modelling the relationships between music and mood have cross-cutting applications in music informatics, from music retrieval and recommendation through content curation and cinematic applications to improved wellbeing (e.g. stress management or enhanced focus) and music therapy. Music mood models to date either focus on a small subset of correlates between musical features and evoked or perceived emotion, or use a data-driven approach to learn and predict associations between music and mood. These models often lack explainability, i.e. the ability to contribute to our understanding of how music impacts human emotion and wellbeing. The aim of this PhD is to develop new methodologies to combine knowledge-driven and data-driven approaches in music mood modelling and validate the approach by creating and testing broadly applicable music mood models either using music retrieval as a use case, or by measuring human physiological response using sensors.
C4DM theme affiliation:
Music Informatics, Music Cognition, Semantic Web, Knowledge Discovery