基于隐含狄利克雷分配的微博推荐模型研究

Micro-blog Recommendation Model Based on the Latent Dirichlet Allocation

微博作为一种新的社交网络,其影响力日益增大,逐渐成为用户获取第一消息的手段。为了增进微博服务,为用户提供兴趣信息,满足用户需求,本文综合考虑微博文本的特殊性,对原有的隐含狄利克雷分配(LDA)模型进行改进,提出基于User-topic的微博推荐模型。该模型在LDA的基础上添加了用户信息,从而得到主题生成用户的概率。通过微博文本-主题-词汇-用户之间的关系为用户提供多角度微博推荐,包括主题推荐、微博内容推荐和用户推荐。

Micro-blog as a new social network has gradually become the main mean of getting the latestnews for users with its high influence. In order to improve the quality of micro-blog＇s service,and provideinteresting information for users, this paper considered the particularity of the micro-blog＇s texts, and im-proved the original LDA model based on User-topic micro-blog recommend model. This model has addedextra user information on the basis of LDA, so as to obtain the user＇s probability for every theme.As a re-sult, we can provide users with multi-angle micro-blog recommendation by analyzing the relationship be-tween the users and text-theme–vocabulary. The content of recommendation includes theme recommen-dation, micro-blog＇s text recommendation, and users themselves recommendation.