An Analysis of Musical Work Datasets and their Current Level of Linkage

Music works are key concepts that present a powerful linkage potential fully acknowledged in the fields of digital music libraries and digital musicology. They form an abstract connecting point for the entities referring to them, and large work datasets act as authority data that offer a promising analysis and search potential. These days, digital music libraries and digital musicology research rely primarily on datasets that have been created over the last decade, mostly from previously existing datasets, such as bibliographic records. In this paper, we try to provide a better understanding of the content of some of the most important datasets available and to evaluate their level of linking. We analyse two leading library datasets, namely those of the Bibliothèque nationale de France (BnF) and the Deutsche Nationalbibliothek (DNB), both available in RDF format, and look at how many works they contain, how these are distributed over time, and their distribution by composer. We compare the results with two other datasets that have completely different backgrounds, namely the Petrucci Music Library (known as IMLSP) and MusicBrainz datasets, two crowd-sourced projects. We evaluate the level of linking the two library datasets currently have with each other through the Virtual International Authority File (VIAF), and their current linking status with other libraries contributing to VIAF. We also evaluate the linking status the IMSLP and the MusicBrainz projects currently have with each other and with other datasets.