The Quest for Ground Truth in Musical Artist Tagging in the Social Web Era

Research in Web music information retrieval traditionally focuses on the classification, clustering or categorizing of music into genres or other subdivisions. However, current community-based web sites provide richer descriptors (i.e. tags) for all kinds of products. Although tags have no well-defined semantics, they have proven to be an effective mechanism to label and retrieve items. Moreover, these tags are community-based and hence give a description of a product through the eyes of a community rather than an expert opinion. In this work we focus on Last.fm, which is currently the largest music community web service. We investigate whether the tagging of artists is consistent with the artist similarities found with collaborative filtering techniques. As the Last.fm data shows to be both consistent and descriptive, we propose a method to use this community-based data to create a ground truth for artist tagging and artist similarity.