基于LDA模型的中文微博话题意见领袖挖掘

有效挖掘微博空间中的话题意见领袖成为亟待解决的热点问题.针对这一问题，提出了基于LDA语义信息和HowNet知识库的短文本子话题分类算法.对分类后的微博从显式、隐式及用户等方面综合衡量微博的影响力，并根据层次分析法对多个因素进行科学地权值分配.实验结果表明，提出的方法较基于支持向量机的方法具有更好的效果，同时提出的影响力度量模型可以有效地挖掘出微博中的话题意见领袖.

Detecting Topical Opinion Leaders Based on LDA Model in Chinese Microblogs

How to effectively detect the opinion leaders in the Chinese microblog space has become a hot research problem in the related area. To tackle this issue, an algorithm combined LDA model and HowNet was proposed to classify the short texts in microblogs based on their underlying subtopics. Then an influence measurement containing the criteria on explicit, implicit and user features was introduced. An analytic hierarchy process method was employed to assign different weight of each parameter. Experiment results showed that the proposed short text classification method outperformed the traditional SVM based method, and the proposed influence measurement model could effectively detect the opinion leaders in the Chinese hot topic microblogs.