Hier-UIM: A hierarchy user interest model for personalized news recommender
Meilian Lu, Jinliang Liu · 2016
User interest model is the key for content based news recommender. However, current user interest modeling methods have some defects, such as do not consider the multi-topic characteristic of news content and news events, resulting in incomplete and inaccurate expression of user interest. Moreover, these methods have the problem that real-time performance is poor when the numbers of users and news increase sharply and cold start problem when the historical data are insufficient. All these will reduce the effects of news recommender. In order to build user interest model more adequately and improve the effect of news recommender, we propose a hierarchy user interest modeling method and corresponding news recommendation method. The experimental results prove that the news recommendation method based on our proposed hierarchy user interest model is superior in precision, and the recommendation results are more diverse and time-sensitive.