一种基于长短期用户表示和多视角学习的新闻推荐方法  

A NEWS RECOMMENDATION METHOD BASED ON LONG-TERM AND SHORT-TERM USER REPRESENTATIONS AND MULTI-VIEW LEARNING

作　　者：何丽[1] 王京豪 He Li;Wang Jinghao(School of Information Science and Technology,North China University of Technology,Beijing 100144,China)

机构地区：[1]北方工业大学信息学院,北京100144

出　　处：《计算机应用与软件》2023年第10期46-53,共8页Computer Applications and Software

基　　金：国家自然科学基金项目(61972003,61672040)。

摘　　要：个性化新闻推荐系统可以帮助用户在海量新闻中快速获取感兴趣内容。用户的兴趣有长期和短期之分,新闻信息也分多种类别,而现有的方法往往基于单类别信息学习新闻的表示。基于此,提出一种融合长短期用户表示、多特征新闻表示的方法。采用基于协同注意力机制的多视角学习方法构建新闻编码器,从新闻的标题、分类和摘要特征中学习统一的新闻表示;利用改进的新闻表示在基于长短期兴趣的用户编码器中进一步细粒度学习用户表示。在真实新闻数据集上的实验结果表明,该方法与其他推荐算法相比在准确率上有明显提高。Personalized news recommendation system can help users quickly get interested content in mass news.The interest of users can be divided into long-term and short-term,and news information can also be divided into many kinds.However,the existing methods often learn news representation based on single kind information.Therefore,this paper proposes a new method which combines long-term and short-term user representation and multi-feature news representation.A multi-view learning method based on cooperative attention mechanism was used to construct news encoder to learn unified news representation from the news title,classification and summary features.The improved news representation was used to further fine-grained learning of user representation in the user encoder based on long-term and short-term user representations.The experimental results on real news data sets show that the proposed method is more accurate than other recommendation algorithms.

关 键 词：长短期用户表示 多视角学习 注意力机制 神经网络 新闻推荐 

分 类 号：TP3[自动化与计算机技术—计算机科学与技术]