Modeling users' heterogeneous taste with diversified attentive user profiles.

Two important challenges in recommender systems are modeling users with heterogeneous taste and providing explainable recommendations. In order to improve our understanding of the users in light of these challenges, we developed the attentive multi-persona collaborative filtering (AMP-CF) model. AMP-CF breaks down the user representation into several latent "personas" (profiles) that identify and discern a user's tastes and inclinations. Then, the exposed personas are used to generate, explain, and diversify the recommendation list. As such, AMP-CF offers a unified solution for both aforementioned challenges. We demonstrate AMP-CF on four collaborative filtering datasets from the domains of movies, music, and video games. We show that AMP-CF is competitive with state-of-the-art models in terms of accuracy while providing additional insights for explanations and diversification. [ABSTRACT FROM AUTHOR]