Recommendation Framework Integrating User Preferences and News Aspect-level Feature Representation
This proposes a news recommendation framework(ALNR)based on aspect-level user preferences and news feature representation learning.By news aspect-level encoders and user aspect-level encoders,it learns user preferences and news aspect-level feature representations separately.The design of these encoders allows for the capture of differences between users and news,enabling the prediction of the probability of a user clicking on a candidate news item,thereby enhancing the personalization level of the recommendation system.Through experiments conducted on the MIND dataset,it has been verified that the proposed method significantly outperforms the current leading baseline method,further demonstrating the importance of aspect-level features.
国家科技期刊平台
NSTL
万方数据
