A1600
Title: Model-based clustering of music pieces
Authors: Yingying Zhang - Western Michigan University (United States)
Xuwen Zhu - The University of Alabama (United States) [presenting]
Hung Tong - The University of Alabama (United States)
Volodymyr Melnykov - The University of Alabama (United States)
Abstract: Grouping songs with similar music compositions plays an important role in streaming services as group characteristics can be used to provide personalized song recommendations and enhance user engagement. In essence, a song is characterized by a categorical sequence consisting of notes from different octaves and the sojourn time of every note. A cluster-weighted model is proposed to cluster songs based on their compositions, with marginal distributions as multivariate first-order Markov models and conditional distributions as mixtures of Gammas. The proposal enjoys great flexibility in capturing potentially multimodal behavior of sojourn time. Application of the model to cluster piano songs achieves meaningful results.
