Music feeling classification based on lyrics using weighting of non-emotional words

Abstract

Music is used by people for music therapy and provision of mental support. Recently, in the field of music information science, studies on music classification have been conducted. However, most of these studies have classified music on the basis of melody and tempo. No classification method that is based on lyrics has been proposed for music therapy and the provision of mental support. Lyrics are an important element of music in addition to melody and tempo. Therefore we propose a music classification method based on lyrics. In this study, we conducted the emotion weighting of non-emotional words by using similarity in order to improve the previously proposed method that uses only emotional words. We realized emotion extraction from the entire lyrics by using non-emotional words in the proposed method. Further, we showed that a relatively high classification precision is obtained when compared with the conventional method.