“…These similarity measures enable applications such as music recommender systems [4,8], automated playlist generators [20,25], or intelligent user interfaces to music collections [23,19]. Computational features for music similarity calculation can be broadly categorized into music content-based, music context-based, and user context-based [34]. While contentbased feature extraction techniques derive the representation of a music item from the audio signal itself [7], music context-based approaches make use of data that are not encoded in the audio signal [30], for instance, the performer's political background, the meaning of a song's lyrics, images of album covers, or co-occurrence information derived from playlists.…”