Global Music Recommendation using Deep Contrastive Learning and Hybridized Filtering

Music is a significant part of any culture’s identity, and has the power to connect people with cultures across the globe. In this paper, we introduce and evaluate a novel Hybridized Filtering approach for Music Recommendation, implemented by employing a Deep Contrastive Learning methodology. We evaluate and compare our proposed methodology to existing and popular recommendation approaches: namely Content-based and Collaborative Filtering. By evaluating the performance metric AURC: Area Under the Recall@K vs K Curve, we conclude that our new approach has a better recommendation performance than the alternatives. We explain how this approach handles and mitigates numerous limitations associated with the others. Furthermore, we demonstrate our model’s ability to provide Global recommendations by recommending songs in a new language (Hindi) for an English user.