基于情感计算的微博突发事件检测方法研究

Bursty Event Detection in Microblogging based on Sentiment Computing

微博已经成为社会新闻舆论最重要的集散地和社会群体平台,对微博信息流进行监测发现,突发事件对于舆情监控具有十分重要的意义。文章通过构建情感向量,采用改进的Kleinberg方法对情感状态进行监测,发现突发事件情感特征及突发期,并采用谱聚类方法对处于突发期的博文进行聚类分析,抽取突发事件。实验结果证明该方法可以快速发现微博流中的突发事件,是一种有效的在线事件检测方法。

Microblogging has become a new social media for public opinion sharing and plays an increasing important role in burst event detection. In this paper, constructed sentiment vectors and built hierarchical structures. Then we monitored the states of sentiment words to discover burst events and burst periods using improved method of Kleinberg. The content of events is formed using spectral clustering approach according the messages in burst periods. Experimental evaluations show that online event detection could implement effectively using the approach proposed in this article.