On weighted hybrid track recommendations

Abstract

Music is a highly subjective domain, which makes it a challenging research area for recommender systems. In this paper, we present our TRecS (Track Recommender System) prototype, a hybrid recommender that blends three different recommender techniques into one score. Since traceability is an important issue for the acceptance of recommender systems by users, we have implemented a detailed explanation feature that supports transparency about the contribution of each sub-recommender for the overall result. To avoid overspecialization, TRecS peppers the result list with recommendations that are based on a serendipity metric. This way, users can benefit from both recommendations aligned with their current taste while gaining some diversification. © 2013 Springer-Verlag Berlin Heidelberg.