Overview of Serendipity in Recommender Systems

Denis Kotkov

Abstract: Has it ever happened to you that services like Spotify, Netflix or YouTube showed you recommendations on the same topic over and over again? This might be caused by the lack of serendipity in recommender systems of these services. Recommender systems are software tools that suggest items, such as audio recordings or videos, of interest to users. Meanwhile, serendipity is the property of these systems, which indicates the degree, to which they suggest items that pleasantly surprise users. In this talk, I will provide an overview of serendipity in recommender systems. In particular, I will talk about how the concept of serendipity has been defined and measured in recommender systems, and what experiments have been conducted to investigate this concept. I will also touch on recommendation algorithms designed to suggest serendipitous items and discuss future directions of the topic.

External IDs:doi:10.1007/978-3-031-62362-2_43