ClusterExplorer: Enable User Control over Related Recommendations via Collaborative Filtering and Clustering
Denis Kotkov, Qian Zhao, Kati Johanna Launis, Mats Neovius · 2020
Related item recommendations have a long history in recommender systems, but they tend to be a static list of similar items with respect to a target item of interest without any support of user control. In this paper, we propose ClusterExplorer, a novel approach for enabling user control over related recommendations. The approach allows users to explore the latent space of user-item interactions through controlling related recommendations. We evaluated ClusterExplorer in the book domain with 42 participants recruited in a public library and found that our approach has higher user satisfaction of browsing items and is more helpful in finding interesting items compared to traditional related item recommendations.