Who likes it more? Mining worth-recommending items from long tails by modeling relative preference

Recommender systems are useful tools that help people to filter and explore massive information. While the accuracy of recommender systems is important, many recent research indicated that focusing merely on accuracy not only is insufficient to meet user needs, but also may be harmful. Other characteristics such as novelty, unexpectedness and diversity should also be taken into consideration. Previous work has shown that more the sales of long-tail items could be more beneficial to both customers and some business models. However, the majority of collaborative filtering approaches tends to recommend popular selling items. In this work, we focus on long-tail item promotion and aggregate diversity enhancement, and propose a novel approach which diversifies the results of recommender systems by considering ''recommendations" as resources to be allocated to the items. Our approach increases the quantity and quality of long-tail item recommendations by adding more variation into the recommendation and maintains a certain level of accuracy simultaneously. The experimental results show that this approach can discover more worth-recommending items from Long Tails and improves user experience.