Browsing Music Recommendation Networks

Many music portals offer the possibility to explore music collections via browsing automatically generated music recommendations. In this paper we argue that such music recommender systems can be transformed into an equivalent recommendation graph. We then analyze the recommendation graph of a real-world content-based music recommender systems to find out if users can really explore the underlying song database by following those recommendations. We find that some songs are not recommended at all and are consequently not reachable via browsing. We then take a first attempt to modify a recommendation network in such a way that the resulting network is better suited to explore the respective music space.