社会媒体用户标签的分析与推荐

Analysis and Recommendation of Social Media User Tags

微博是Web2.0时代重要的网络服务工具,作为以用户为中心的信息发布、传播和分享平台,它包含了非常丰富的用户信息。在微博中,可以使用标签表示用户的兴趣和属性。而一个用户的兴趣和属性,通常包含在这个用户的文本信息和网络信息中。针对微博用户的标签进行分析,提出网络正则化的标签分发模型(NTDM)来为用户推荐标签。NTDM模型对用户个人简介中的词语和标签之间的关系进行建模,同时利用其社交网络结构作为模型的正则化因子。在真实数据集上的实验表明,NTDM在效果以及效率上都优于其他方法。

Microblog is an important online service in Web 2.0. As a platform for web users to post messages, communicate and share information, microblog contains rich information of users. Microblog services can use tags to represent interests and attributes of users. Meanwhile, the interests and attributes of a microblog user also hide behind his/her text and network information. In this paper, we quantitatively analyze user tags and propose a network- regularized tag dispatch model for user tag recommendation. NTDM models the semantic relations between words in user descriptions and tags, with social network structure as its regularization factor. Experiment results in a real world dataset show its effectiveness compared to other baseline methods.