September 27, 2017

Eugenio Tacchini: Handling eclectic tastes in recommender systems: novelty, serendipity and mentors


Recommender systems sometimes fail in recommending the right content to users having eclectic tastes; they can, especially during the first interactions, incur in over-specialization and popularity bias problems. We are investigating on how this problem can be handled and in particular on how to effectively induce novelty and serendipity in recommendations in order to increase user satisfaction.