Exploring Coclustering for Serendipity Improvement in Content-Based Recommendation

Content-based recommender systems are widely used in different domains. However, they are usually inefficient to produce serendipitous recommendations. A recommendation is serendipitous if it is both relevant and unexpected. The literature indicates that one possibility of achieving serendipity in recommendations is to design them using partial similarities between items. From such intuition, coclustering can be explored to offer serendipitous recommendations to users. In this paper, we propose a coclustering-based approach to implement content-based recommendations. Experiments carried out on the MovieLens 2K dataset show that our approach is competitive in terms of serendipity.