An Analysis on Time- and Session-aware Diversification in Recommender Systems

In modern recommender systems, diversity has been widely acknowledged as an important factor to improve user experience and, more recently, intent-aware approaches to diversification have been proposed to provide the user with a list of recommendations covering different aspects of her behavior. In this paper, we propose and analyze the performances of two diversification methods taking into account temporal aspects of the user profile: in the first one we adopt a temporal decay function to emphasize the importance of more recent items in the user profile while in the second one we perform an evaluation based on the identification and analysis of temporal sessions. The two proposed methods have been implemented as temporal variants of the well-known xQuAD framework. In both cases, experimental results on Netflix 100M show an improvement in terms of accuracy-diversity balance.