Modeling the complexity of music metadata in semantic graphs for exploration and discovery

Representing and retrieving fine-grained information related to something as complex as music composition, recording and performance is a challenging activity. OEis complexity requires that the data model enables to describe di,erent outcomes of the creative process, from the writing of the score, to its performance and publishing. In this paper, we show how we design the DOREMUS ontology as an extension of the FRBRoo model in order to represent music metadata coming from di,erent libraries and cultural institutions and how we publish this data as RDF graphs. We designed and re-used several controlled vocabularies that provide common identi€ers that overcome the di,erences in language and alternative forms of needed concepts. OEese graphs are interlinked to each other and to external resources on the Web of Data. We show how these graphs can be walked through for designing a web-based application providing an exploratory search engine for presenting complex music metadata to the end-user. Finally, we demonstrate how this model and this exploratory application is suitable for answering non-trivial questions collected from experts and is a €rst step towards a fully NLGedged recommendation engine.