Is there a data model in music notation?

Scores are structured objects, and we can therefore envisage operations that change the structure of a score, combine several scores, and produce new score instances from some pre-existing material. Current score encodings, however, are designed for rendering and exchange purposes, and cannot directly be exploited as instances of a clear data model supporting algebraic manipulations. We propose an approach that leverages a music content model hidden in score notation, and define a set of composable operations to derive new “scores” from a corpus of existing ones. We show that this approach supplies a high-level tool to express common, useful applications, and can easily be implemented on top of standard components.