Collaborative Music Similarity and Recommendation

Abstract

In this chapter, we illustrate how the data on interaction traces between users and music items, whose acquisition we discussed in the previous chapter, can be used for the tasks of determining similarities between music items and between users. We particularly focus on the topic of music recommendation since it is presumably the most popular task carried out on this kind of data. We first start with a discussion of methods to infer music similarity via co-occurrence analysis, i.e., define similarity via information on which items are listened to together by users (Sect. 8.1). Subsequently, Sect. 8.2 shows how to exploit the graph structure of artist and of user networks, who are connected on various social media platforms.