Bridging the information gap between structural and note-level musical datasets

While there are an increasing number of datasets containing various features of musical information, the lack of connections between them remains a barrier to their use in research. For example, one dataset might encode the identification of structural segments by musicologists in audio recordings, while another dataset could contain a symbolic encoding of the music notation being played in that audio recording. Without explicit connections, there is a significant extra work in realizing their potential for musicological study. In this paper we investigate how Linked Data can be used to implement such connections, specifically between the McGill Billboard corpus of structural annotations and the MIDI Linked Data Cloud (MIDI-LD). Firstly, we republish structural information from Billboard as RDF. We then align this structural data with matching symbolic encodings in MIDI-LD; before finally linking individual structural annotations from Billboard to note-level sections in the MIDI-LD. Our alignments enable cross-referencing and combined queries for musicological analysis across the enriched union dataset, and serve as a model for the creation of information resources comprising musical structures at varying granularity