面向LDA和VSM模型的微博热点话题发现研究  

Research of Micro-Blog Hot Topic Detection Based on LDA and VSM Model

作　　者：李勇[1] 张克亮[1] 

机构地区：[1]解放军外国语学院,河南洛阳471003

出　　处：《自动化技术与应用》2016年第8期52-57,共6页Techniques of Automation and Applications

摘　　要：针对现有话题检测技术的不足,提出了一套适用于微博的热点话题发现方法。通过分析话题检测和微博的相关概念、特点及传播规律,对微博进行预处理和特征项的选择。利用VSM和LDA模型对其进行混合建模并进行微博相似度计算,融合微博社会关系提出了SPWSR聚类算法进行热点话题发现。实验结果表明,在NIST的评价指标体系下,该方法各指标平均提高了10%到20%。Aiming at the shortcomings of the existing topic detection technology, this article puts forward a set of suitable for Micro-blog hot topic detection method. Through the analysis of the relevant concepts of topic detection and Micro-blog characteristics and spread rule, Micro-blog is preprocessed and feature is selected. By using VSM and LDA model for hybrid modeling and feature weighting computation, proposing SPWSR algorithm combined social relations of Micro-blog document is used to hot topic detection. The experimental results show that under the evaluation index system of NIST, each index of the method increase by 10% to 20% on average.

关 键 词：微博 向量空间模型 LDA模型 话题发现 社会关系 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]