The DCML corpus initiative currently focuses on the collection of high-quality MuseScore 3 files that are being annotated right within the engraving software by our annotators. In order to evaluate this data, i.e. annotation labels, notes, and score markup (such as articulation, lyrics, or dynamics) it needs to be extracted from the engraving files and brought into an accessible and interoperable format. This is where the parsing library ms3 comes into play that is being developed in Python. It exposes a command-line interface that lets you, for example, easily extract score information and store it as TSV tables, and plays a crucial part in the DCML annotation workflow. Although it can be used for any kind of custom labels, it has hard-coded functionality for dealing with the DCML harmony annotation standard. More than that, the library can be used to load in such TSV files to transform and evaluate them.
