LSTM을 활용한 이별 가사 감정 분류 모델 연구

Abstract

With the development of the Internet and smart phones, the productivity of digital sound sources has improved, making it easier to access, and as a result, users are increasingly interested in music search and recommendation. The existing music recommendation system is a method of recommending music according to the user's situation, mood, and research has been conducted to classify emotions using only melodies such as pitch, tempo, and beat. However, in music, lyrics are one of the means of expressing human emotions, and the role is gradually increasing, and development has been made to expand the study of combining music and natural language processing (NLP). In addition, Naver 'Vibe' platform is based on audio-based music recommendation and content-based recommendation using deep learning. In this study, we propose a parting lyrics emotion classification model that recommends lyrics of similar emotions by embedding the words of words through Word2Vec learning and recommending lyrics of similar emotions for recommending music that segmented emotions of farewell.