图书情报工作 ›› 2019, Vol. 63 ›› Issue (4): 6-14.DOI: 10.13266/j.issn.0252-3116.2019.04.001

• 专题:社交媒体网络舆情的情感分析及管控策略研究 • 上一篇    下一篇

基于模块度的话题发现及网民情感波动研究——以新浪微博“中美间贸易摩擦”话题为例

张海涛1,2, 刘雅姝1, 张枭慧1, 宋拓1   

  1. 1. 吉林大学管理学院 长春 130022;
    2. 吉林大学信息资源研究中心 长春 130022
  • 收稿日期:2018-05-16 修回日期:2018-09-26 出版日期:2019-02-20 发布日期:2019-02-20
  • 通讯作者: 刘雅姝(ORCID:0000-0003-0742-5901),博士研究生,通讯作者,E-mail:1911670944@qq.com
  • 作者简介:张海涛(ORCID:0000-0002-9421-8187),教授,博士生导师;张枭慧(ORCID:0000-0003-0020-3126),硕士研究生;宋拓(ORCID:0000-0003-1282-1861),博士研究生。

Research on Topic Discovery Based on Modularity and Sentiment Fluctuation of Internet Users——Taking Sina Weibo's “China-US Trade Friction” as an Example

Zhang Haitao1,2, Liu Yashu1, Zhang Xiaohui1, Song Tong1   

  1. 1. Management School of Jilin University, Changchun 130022;
    2. The Information Resource Research Center of Jilin University, Changchun 130022
  • Received:2018-05-16 Revised:2018-09-26 Online:2019-02-20 Published:2019-02-20

摘要: [目的/意义]探索热点事件评论网络中话题社群及网民的情感波动,掌握舆情事件发展过程,对于整体把握热点事件的发展方向,做好新时期网络舆论的引导工作具有重大意义。[方法/过程]以复杂网络理论为基础,基于评论词语间的共现关系构建基于事件发展的子事件网络,通过社群发现算法来识别子事件评论网络中的话题社群,将情感词依据情感词典赋予情感分类属性,基于事件的演化过程动态地跟踪网民意见以及情感波动。[结果/结论]研究结果表明,评论网络群落发现以及变异系数方法可以有效地衡量网民话题讨论的规模与集中程度;评论网络中赋予情感词节点情感分类属性方法可以体现事件演化过程中网民的情感变化;舆论衍生话题对事件的舆情发展有持续性影响;网民话题讨论内容对于事件演化具有一定程度上的前瞻性。

关键词: 复杂网络, 社群发现, 话题发现, 情感分析

Abstract: [Purpose/significance] Exploring topical communities and sentiment fluctuations of Internet users and grasping the process of development of events have great significance to control the development direction of the events and lead guidance of the network public opinion in the new period. [Method/process] Based on the theory of complex networks, the study constructed sub event network based on co-occurrence relations among user comments, identifying topic community in sub-event commenting networks through community discovery algorithms and giving the attribute to emotion word according to the emotional dictionary. The study dynamically tracked the opinions and emotions of Internet users based on the evolution process of events. [Result/conclusion] The conclusion showed that the commenting network community discovery and coefficient of variation method can effectively measure the scale and concentration of Internet users' topic discussion; emotional word sentiment classification attribute method can reflect the emotional changes of Internet users in the process of event evolution; the derived topic of public opinion has a continuous influence on the development of the event public opinion; the content of the topic discussion of Internet users has some foresight to the evolution of the event.

Key words: complex network, community discovery, topic discovery, sentiment analysis

中图分类号: