融合用户兴趣模型与会话抽取的微博推荐方法

Microblog recommendation method based on user interest model and conversation extraction scheme

针对传统推荐方法在短文本处理方面的不足,提出一种基于用户兴趣模型与会话抽取算法的微博推荐方法。该方法应用基于归一化割加权NMF的微博用户兴趣模型获取用户—主题矩阵,产生用户感兴趣的微博主题,结合基于Single-Pass聚类模型的会话在线抽取算法SPFC(single-pass based on frequency and correlation)获取微博的会话队列,并与用户感兴趣的微博主题进行相似度计算,最后得到实时的微博推荐结果。实验表明,此方法能有效地进行微博推荐。

Aiming at the deficiency of conventional recommendation method in short text message processing, this paper pro- posed a microblog recommendation method based on user interest model and conversation extraction. First, it applied a Ncut weighted non-negative matrix factorization （Ncut weighted NMF）to obtain user-interest matrix. And then used Single-Pass clustering based on frequency and correlation for conversation extraction to obtain mieroblog conversation. Experiments show that this method can effectively cluster micro-blogs and support micro-blog recommendation.