A News Recommendation Approach Based on the Fusion of Attention Mechanism and User's Long and Short Term Preferences

Nowadays, more and more news readers are reading news online and they have access to millions of news articles from multiple sources. To help users find correct and relevant content, news recommendation systems have been developed to alleviate the problem of information overload and to recommend news items that may be of interest to news readers. Compared to traditional recommendation domains such as product recommendation and video recommendation, in the news recommendation domain, the update speed of information is very fast, so when recommending news content to users, it is necessary to consider the deeper characteristics of the user's representation and the recommendation speed of the news, which in turn improves the accuracy of the news recommendation user satisfaction, and increases the user's stickiness. Therefore, this paper proposes a news recommendation method that integrates the attention mechanism with the user's short-term and long-term preferences, which breaks through the problem that the news feature extraction does not make full use of the full text and the modeling of the user's interest is not granular enough and timely, and at the same time, while achieving a high level of accuracy, it also meets the requirements of high timeliness.