计算机科学
图形
信息过载
情报检索
推荐系统
钥匙(锁)
任务(项目管理)
注意力网络
万维网
人工智能
理论计算机科学
计算机安全
管理
经济
作者
Yong Xu,Bao Chen,Jingru Zhen,Guangfu Ma,Gongbin Chen,Yan Liu,Qun Fang
标识
DOI:10.1145/3562007.3562030
摘要
News recommendation is necessary to help users find interesting news, improve their experience, and alleviate information overload. Accurately learning news and user representations is a key task in news recommendation systems. News texts usually contain rich entities, however existing recommender systems ignore the importance of news entities. In order to effectively alleviate the above problems, we design a multi-view news recommendation system based on knowledge graph. First, with news headlines, summaries, categories, and knowledge graph features, we learn news representations using a graph interactive attention network and a multi-head attention mechanism. Second, we combine a recurrent neural network and an interactive attention network to learn user representations from user historical click news records. Finally, predict the probability that the user will click on the candidate news. This method effectively alleviates the problem that the current news recommendation model has a weak ability to capture news representations and user interest representations. Experiments on real datasets show that this method can effectively improve the performance of news recommendation.
科研通智能强力驱动
Strongly Powered by AbleSci AI