Integrando Avaliações Textuais de Usuários em Recomendação baseada em Aprendizado por Reforço

In Multi-Armed-Bandit (MAB) approaches for Recommendation Systems, items are represented as arms to be recommended and the goal is to maximize the expected user’s satisfaction (i.e., reward). Despite the reward often being the ratings explicitly assigned by the user, in other scenarios, implicit ratings extracted from user comments by review-aware recommendation systems (RARs) may efficiently elucidate the user’s preferences. In this paper, we provide a preliminary study of the impact of using these implicit ratings instead of explicit ones in MAB approaches. Our results point out that implicit ratings decrease the entropy of the datasets, negatively impacting the performance of MAB.