A tag-level factor graph model for semantic music discovery

This paper proposes a semantic music discovery system based on a tag-level factor graph (TFG) model with utilization of tag probability and content similarity in a unified fashion. The content similarities are calculated based on the extracted pitch features while tag probabilities are obtained from our previous auto-tagging system. The TFG model consists of a set of node and edge feature functions, which define the impact of tag probabilities and content similarities on representative degrees of songs. Representative degrees indicate to which extent a song is a representative one given the query tag. The loopy max-product inference algorithm is applied to obtain the values of all representative degrees that maximize the joint probability distribution of the TFG model. Experiment results show the TFG model improves the performance by 5.6% higher in the precision rate at top 3 music and 3.5% higher at both top 5 and 10 music.