Extracting patterns from tracks using emotional tags
Michael Kai Petersen, Andrius Butkus · 2008
Abstract. Despite the idiosyncratic character of tags applied to songs in social networks like last.fm, recent studies have revealed that users often tend to agree on the affective terms they attach to music. Using some of these frequently occurring words as emotional buoys to form a semantic plane of psychological valence and arousal dimensions, we project lyrics into this space and apply LSA latent semantic analysis to model the affective context of a number of songs. We compare the components retrieved from the lyrics with the user defined affective terms that constitute the tag clouds of the corresponding songs at last.fm, and propose that LSA could be applied to extract structural patterns as a basis for automatically generating emotional playlists. 1