simREC

The joint project aims at developing a hybrid recommendation system. In a first phase a standard collaborative filtering approach for millions of user-generated data (implicit feedback while using a music streaming service) is developed. In parallel a content-based approach using semantic metadata and the underlying semantic graph structure is explored (Freebase, Musicbrainz). In a second phase both components are merged by late-fusion and we will try different hybridisation strategies which will be evaluated within explanation interfaces with real-world users.