SONG TOPIC INFERENCE INFORMATION GENERATION SYSTEM

摘要

Provided is a song topic inference information generation system that is able to provide information that is more effective than the prior art in interpreting the topic of a song. A topic number learning means 9 executes for a prescribed number of times a topic number update learning calculation in which a topic number update calculation is performed for all of multiple items of song data for each of multiple artists. A random number generator having a bias for an occurrence probability corresponding to the probability distribution of the topic numbers is used to perform the topic number update calculation, in which a topic number assigned to a given item of song data S_ar for a given artist a is updated. On the basis of the learning result from the topic number learning means 9, an output means 13 outputs, for each of the multiple artists, a topic number for each of the items of song data and a word probability distribution for each topic number.