Enhancing Music Discovery: A Real-Time Recommendation System using Sentiment Analysis and Emotional Matching with Spotify Integration

This paper presents an emotion-based music recommendation system that utilizes sentiment analysis and emotion recognition to provide personalized music suggestions based on the user's emotional state. The system processes both text and audio inputs, using emotion detection models such as Hugging Face's aiknowyou/it-emotion-analyzer and pysentimiento for sentiment classification. Music recommendations are made by mapping the detected emotions to relevant music genres with the help of the Spotify API. The system was evaluated based on emotion detection accuracy and the relevance of the music recommendations. Results showed high accuracy in emotion detection from text inputs, with text achieving an accuracy rate of 92.5% for joy and 94.3% for sadness, while audio inputs demonstrated slightly lower accuracy, with a maximum of 91.0% for sadness. The music recommendation system achieved an 88.5% user satisfaction rate, with over 90% accuracy in artist-specific matching. Despite the challenges in mapping emotions like fear and surprise to appropriate genres, the system provided fast real-time recommendations with an average response time of 350 milliseconds. Overall, the proposed system offers a personalized and emotionally responsive music recommendation experience, with areas for future improvement in audio input recognition and emotion-to-genre mapping.