Over the past century, sociocultural and technological developments have fostered the emergence of what Peterson and Kern (1996) call “omnivorous” music listeners, as well as non-hierarchical forms of categorization like tagging. Despite such trends, genre remains the primary basis for ascertaining similarity in systems with musical content, metadata, or both. Furthermore, techniques employed within many recommender systems indirectly continue to reflect genre-based categorization and taste. This paper will provide an overview of the contexts in which such trends and tensions have emerged. It will also consider prospects for incorporating more actively nuanced dimensions of similarity into recommender systems, which could enable users to engage more easily in cross-genre music discovery. To provide further grounding for such possibilities, the paper will discuss preliminary findings from a study that employs semi-structured interviews and music-seeking exercises conducted with “avid recreational music listeners,” with a focus on the ways they conceptualize musical similarity.
- Recommender Systems,
Available at: http://works.bepress.com/jason_neal/8/