Presentation Information
[1H4-OS-5a-06]Extracting Microstructure of Music Genres through Playlist Co-occurrence Network Analysis
〇Makoto Takeuchi1 (1. CyberAgent, Inc.)
Keywords:
Music genres,Network analysis,Community detection,Playlists,Co-occurrence network
In the streaming era of music consumption, "microgenres" that cannot be captured by traditional macro-genre classifications (Rock, Rap, Pop, etc.) are gaining importance. However, the definition of microgenres is often discussed intuitively, and there is insufficient discussion of clear definitions. This study proposes a method to structurally define microgenres from community structures in playlist co-occurrence networks. We used data from playlists created by users of music streaming services. We performed community detection on the artist playlist co-occurrence network and extracted communities. To measure the "microgenre-ness" of each community, we developed a Conductance Deficit index based on comparison with the Configuration Model. This index measures "how much stronger the boundary is compared to random" and has the advantage of being independent of community size. This approach presents a promising method for extracting fine-grained structures without relying on fixed labels and provides a new perspective on the problems with macro-genre labels pointed out in previous research.