基于微博分类的用户兴趣识别

Identifying User Interests based on Microblog Classification

社会媒体成为用户分享与获取信息的重要平台。发现感兴趣的微博账户与信息是社交媒体平台最重要的活动,其关键问题在于用户兴趣模型的构建。提出基于微博分类的用户兴趣识别方法。首先人工构建目标分类体系,基于典型微博账户采集微博训练语料训练微博分类器,而后通过对用户微博进行分类识别出用户感兴趣的类别。实验表明基于典型主题类别微博,结合词语与主题的特征可有效进行微博分类达到86%的F值,输出的类别可准确表示用户兴趣。

Social media （such as Sina microblog） becomes important platform for sharing and accessing information for users. Identifying potential interested weibo accounts and microblogs are the most important tasks on social media. The key challenge is user modeling. This paper proposes a user modeling method based on microblog classification. First,construct target taxonomy, collect training data from typical weibo accounts for training the microblog classifier, then identify user interests by classifying the microblogs of the users. The experiments show that based on the collected training data, the combination of word features and topic features is effective for classifying microblogs by achieving 86% F - measure and the output categories represent user interests accurately.