Wav2vec-Based Intelligent Music Recommendation from Text Emotions

Emotion-based music recommendation systems analyze the user’s emotional state to recommend music. They apply artificial intelligence to suggest music that influences the user’s mood. In this paper, the authors consider the problem of recommending music to users in a mental state of depression. Depression is detected in user’s texts using natural language processing techniques. Upon detecting depression, the proposed system suggests music in appropriate ragas that can have a therapeutic influence on the mental state. The paper proposes using a contextual deep learning model, Wav2vec to identify the required ragas alongside the conventional audio features. The proposed method achieves state-of-the-art performance in raga identification, suggesting the significance of such models in other raga classification problems.