Listen To Me - Don't Listen To Me: What Communities of Critics Tell Us About Music

Social knowledge and data sharing on the Web takes many forms. So too do the ways people share ideas and opinions. In this paper we examine one such emerging form: the amateur critic. In particular, we examine genius.com, a website which allows its users to annotate and explain the meaning of segments of lyrics in music and other written works. We describe a novel dataset of approximately 700,000 users’ activity on genius.com, their social connections, and song annotation activity. The dataset encompasses over 120,000 songs, with more than 3 million unique annotations. Using this dataset, we model overlap in interest or expertise through the proxy of co-annotation. This is the basis for a complex network model of the activity on genius.com, which is then used for community detection. We introduce a new measure of network community activity: community skew. Through this analysis we draw a comparison of between co-annotation and notions of genre and categorisation in music. We show a new view on the social constructs of genre in music.