P4-15: A Dataset of Symbolic Texture Annotations in Mozart Piano Sonatas
Couturier, Louis*, Bigo, Louis, Leve, Florence
Subjects (starting with primary): Evaluation, datasets, and reproducibility -> novel datasets and use cases ; Musical features and properties ; MIR fundamentals and methodology -> symbolic music processing ; Domain knowledge -> computational music theory and musicology ; MIR tasks -> music transcription and annotation
Presented In-person, in Bengaluru: 4-minute short-format presentation
Musical scores are generally analyzed under different aspects, notably melody, harmony, rhythm, but also through their texture, although this last concept is arguably more delicate to formalize. Symbolic texture depicts how sounding components are organized in the score. It outlines the density of elements, their heterogeneity, role and interactions. In this paper, we release a set of manual annotations for each bar of 9 movements among early piano sonatas by W. A. Mozart, totaling 1164 labels that follow a syntax dedicated to piano score texture. A quantitative analysis of the annotations highlights some characteristic textural features in the corpus. In addition, we present and release the implementation of low-level descriptors of symbolic texture. These descriptors can be correlated with texture annotations and used in different machine-learning tasks. Along with provided data, they offer promising applications in computer assisted music analysis and composition.