Mixing Music as Linked Data: SPARQL-based MIDI Mashups

A large number of datasets about music are available today in the Linked Open Data cloud, but most of them only describe music metadata. Datasets representing music notation (i.e. fine-grained musical transcriptions) are scarce, and hence musicians do not have the possibility to exploit Web technologies to their full potential. In particular, this situation hampers the musician’s process of creating mashups, new musical compositions produced by remixing existing tracks. Recently, the MIDI Linked Data cloud has interlinked and published more than 300K MIDI songs as Linked Data. In this paper, we investigate the use of Semantic Web technology to produce musical mashups, and we present a framework to generate them systematically. We evaluate our approach with SPARQL-DJ, a prototype implementation that can be used to find, match, select and synchronize existing MIDI Linked Data, mix them, and create new musical content.