Research Paper Recommender System with Serendipity Using Tweets vs. Diversification

Abstract

So far, a lot of works have studied research paper recommender systems. However, most of them have focused only on the accuracy and ignored the serendipity, which is an important aspect for user satisfaction. The serendipity is concerned with the novelty of recommendations and to which extent recommendations positively surprise users. In this paper, we investigate a research paper recommender system focusing on serendipity. In particular, we examine (1) whether a user’s tweets lead to a generation of serendipitous recommendations and (2) whether the use of diversification on a recommendation list improves serendipity. We have conducted an online experiment with 22 subjects in the domain of computer science. The result of our experiment shows that tweets do not improve the serendipity, despite their heterogeneous nature. However, diversification delivers serendipitous research papers that cannot be generated by a traditional strategy.