Flexible Querying of Musical Score Databases
Soutenue par Adel ALY, à l’Université de Rennes (IRISA – ENSSAT), le 02/12/2025. Dirigée par Virginie Thion. Excellent travail, avec des idées que j’aurai aimé mettre en oeuvre après mes travaux avec Philippe Rigaux sur la notation musicale. Toujours un plaisir de partager des moments avec Virginie et François. Un régal pour les oreilles avec des partitions de musiques bretonnes qui me rappellent mon grand-père [theses.fr] [Rapporteur]
This thesis addresses the repre- sentation and querying of symbolic musical scores from a database perspective, with a focus on flexible melodic pattern retrieval. Existing approaches fail to capture musically relevant variations and local deviations in pitch and duration. To address this limitation, we propose a framework for flexible querying grounded in an algebraic model of musical data, where scores are represented as time-based structures and queried through selection operators. Flexibility is introduced through fuzzy mechanisms supporting note-level variations and global transformations, polyphonic constraints, and contour-based abstractions. The framework is implemented using a prop- erty graph model and an extension of the Cypher query language. Queries are compiled into executable graph queries, and results are evaluated and ranked. The system is integrated into SKRID, a digital library of Breton traditional music. Experiments show that the approach remains tractable, with execution times scaling linearly with pattern size and database size, and with limited overhead due to flexible matching. This work provides both a formal foundation and a practical system for flexible querying of symbolic musical score databases.