HEMR:Hypergraph Embeddings for Music Recommendation

With the increasing number of multimedia streaming platforms, it has become essential to provide advanced recommendation systems to support users in browsing the vast amount of multimedia data according to their preferences and needs. The key challenge is to model entities and their complex relationships, such as users’ listening patterns, song features, and artists’ releases. This paper represents an extended abstract of a recent work describing of a novel framework, namely Hypergraph Embeddings for Music Recommendation (HEMR), which leverages hypergraph data structures along with modern graph machine learning techniques for song recommendation. The hypergraph data model is used to represent complex interactions between users and songs, while embedding techniques help to infer user-song similarities via a vector mapping. Our experiments demonstrate HEMR’s effectiveness and efficiency compared to state-of-the-art music recommender systems, especially in cold start problem scenarios. Therefore, our system is a promising solution to embed within a music streaming platform to enhance users’ satisfaction.