基于LDA主题模型的微博标签生成研究  

Microblog Tags Generation Based on LDA Theme Model

作　　者：熊回香[1] 叶佳鑫 XIONG Hui-xiang, YE Jia-xin(School o fin formation Management, Central China Normal University, Wuhan 430079, China)

机构地区：[1]华中师范大学信息管理学院,湖北武汉430079

出　　处：《情报科学》2018年第10期7-12,共6页Information Science

基　　金：国家社会科学基金项目(12BTQ038)

摘　　要：【目的/意义】为微博用户生成质量较高的标签,帮助用户对微博特征进行描述。【方法/过程】首先,分析微博的特征建立候选标签与候选词表。随后,用LDA对用户的微博进行主题分析,将用户的微博与预选标签相匹配,生成预选标签。最后,用LDA对用户关注的人及用户的粉丝的微博进行主题分析,对预选标签进行控制生成最终标签。【结果/结论】实验结果表明,本文以LDA为基础提出的标签生成方法具有不错的效果,最终生成的标签能够较为准确地描述用户的微博特征。【Purpose/significance】To generate high quality tags for microblog＇s user to help user to describe the characteristics of microblog.【Method/process】Firstly, to analyze the characteristics of microblog and build the candidate tags and candidate word table.Then, analyze the user＇s microblog topics by LDA and the user＇s microblog was matched with the primary tag to generate the primary tags.Finally,the topics of user＇s followers and user＇s attentions were analyzed by LDA to control the primary tags and generate the final tags.【Result/conclusion】The experimental results show that the tag generation method based on LDA has a good effect, and the resulting tag can accurately describe the user＇s microblog characteristics.

关 键 词：微博 标签生成 标签质量 LDA 

分 类 号：G206.3[文化科学—传播学]