基于友邻-用户模型的微博主题推荐研究

在微博社交网络中，微博文本内容短小，主题覆盖较少，同时主题变化快，用户兴趣更新频繁。已有用户模型不能完全准确刻画微博用户变化的兴趣。友邻集由用户认知度高的群体组成，友邻集的主题兴趣可以全面反映目标用户的多样化兴趣。利用目标用户的友邻集，在本体用户模型上构建微博用户的友邻主题兴趣集，计算更新友邻主题兴趣度，提出友邻-用户模型的实现算法。实验表明，在微博社交网络平台中，友邻一用户模型的微博主题推荐精度要优于传统的用户模型。

Research on themes recommendation in micro-blogging scenario based on neighbor-user profile

In micro-blogging social network, micro-blog content is short text and covered less themes. In addition, themes in network changed fast and user＇s interests updated frequently. Existing user profile can＇t accurately depict vari- ous interests of user. Group of users with high awareness can form neighbor set of user, whose topic interests can com- prehensively reflect diverse interests of target user. A neighbor-user profile implementing algorithm was proposed by building neighbor theme interest set of ontology user profile and calculating update of neighbor theme interest degree of ontology user profile in terms of neighbor set of target user. Experiments show that the accuracy of micro-blogging themes recommendation based on neighbor-user profile performs better than that of traditional individual user profile in the micro-blogging social network platform.