Can a rock song have a jazz audience? : relationship between folksonomy and collaborative filtering in music recommender systems

In this paper we investigate the relationship between a folksonomy-based music classification and a music classification based on collaborative filtering, i.e. on the users' listening behavior. We found a correlation between folksonomy-based songs clustering and clustering computed using methods based on the audience listening behaviour and, using a combination of the two approaches, we also computed the eclecticism level of a sample set of users, finding that eclecticism seems to be a characteristic which changes according to the genre of music most loved by a user.