Fig. 1

Overview of our data transformation workflow, generalised for arbitrary music annotations, and used here for chord and key annotations prior to the generation of the ChoCo Knowledge Graph. The JAMifier ingests chord collections (where metadata and music annotations follow collection-specific conventions and formats) to generate a JAMS dataset. This achieves two integration levels, as all metadata are consistently re-organised, and the music annotations (i.e. chord progressions, in this case) are all encoded and stored in separate JAMS files – one per track/score. The Chonverter achieves notational interoperability among collections by converting the original annotations to the same notational families. Finally, jams2rdf leverages notation-specific ontologies to generate RDF triples and create a Music Knowledge Graph.