Energy-Valence 모델 적용을 통한 문장 감성 기반 음악 추천 시스템 구현

Implementation of Text Emotion-based Music Recommendation System Employing Energy-Valence Model

In this study, we present Feelic, an emotion-based music recommendation application that analyzes users' emotions and recommends customized music. The system classifies the user's diary into five emotions: joy, sadness, anger, anxiety, and neutrality using the KoBERT model, and recommends music that matches the emotions by extracting audio characteristics of various songs using Spotify API. Through K-means clustering, music is divided into 30 clusters and songs belonging to clusters such as user-preferred music are recommended. The application is designed so that users can easily write a diary, analyze emotions, and recommend music accordingly, and aims to manage emotions and improve mental health. To prove the usefulness and effectiveness of Feelic, an emotion analysis system, a music recommendation system, and an application implementation process are described in detail. The results are evaluated with the energy-valence model and presented as reasonable.