A Web-based Approach to Determine the Origin of an Artist

One can define the origin of an artist as the geographical location where he started his career. The origin is an important metadata element, because it can help to specify subgenres, be an indicator of regional popularity and improve recommendations. In this paper, we present six methods to determine the origin, based on Web data sources: one extracts data from Last.fm, two query Freebase and three analyze biographies. We evaluate the different methods with 11275 artists. Circa 55% of the artists can be classified using biographies. The best Freebase method can classify 26% and the Last.fm based method 7%. When comparing on accuracy, the Last.fm and Freebase methods perform similarly with around 90% accuracy. For the biography-based methods we achieve 71%. To improve coverage, a final, hybrid method achieves 77% accuracy and 60% coverage. The accuracy of the continent classification is 87%. As a showcase for our classifier, we developed a mashup application that displays, among others, information about the origin of artists from radio station playlists on a map.