基于LDA模型的网络舆情事件话题演化分析

摘要

通过挖掘网络舆情事件话题演化，有助于在事件全面爆发之前的更早阶段釆取应急措施。本文基于网络舆情事件时间性强、话题与时间关联度单一的特征，选择网民关注度高的食品安全网络舆情事件新闻报道建立文本集，基于LDA模型抽取话题，使用后离散时间型话题模型思路分析话题热度变化，用先离散时间型话题模型思路分析话题内容迁移。实验表明，此思路能够较全面体现话题演化路径，为网络舆情事件分析提供有效途径。%By exploring the event topic evolution of network public opinions, it may help us to take effective measures timely in dealing with emergency cases. Since network public opinion of event is time-sensitive and the topic is closely associated with time, we selected network public opinion of food safety incidents with high concern among netusers and extracted topics based on LDA model. The time post-discretized method was used to analyze the change of topic heat level and the time pre-discretized method was applied to analyze the change of topic content. The proposed method was experimentally verified to be efficient for detecting the event topic evolution of network public opinions.