Improving recommendation lists through topic diversification

摘要: In this work we present topic diversification, a novel method designed to balance and diversify personalized recommendation lists in order reflect the user's complete spectrum of interests. Though being detrimental average accuracy, show that our improves user satisfaction with lists, particular for generated using common item-based collaborative filtering algorithm.Our builds upon prior research on recommender systems, looking at properties as entities their own right rather than specifically focusing accuracy individual recommendations. We introduce intra-list similarity metric assess topical diversity diversification approach decreasing similarity. evaluate book data, including offline analysis 361, !, 349 ratings an online study involving more 2, 100 subjects.