面向话题的中文微博观点倾向性分析研究

Topic Oriented Sentimental Analysis on Microblog

随着微博快速崛起,每天数以千万的人通过微博分享自己对各类话题的观点与情感,如何自动感知微博社区对特定话题的观点倾向性,已经成为中文微博计算亟待解决的问题。由于微博内容短小且不规范,传统的情感分析效率低下且效果很难满足实际需求。现提出一种将情感词典分类的方法进行实验研究,针对腾讯微博20个话题约17 500条微博32 000个句子的数据进行实验,实验结果表明提出的情感词典分类方法效果很好。

With the rapid development of microblogs, millions of users share their opinions and sentiments on different topics by mierobloging every day. Automatically analyzing users＇ sentimental trends and determining the tendency of the whole microblog community under certain topic, have been the basic urgent scientific problem in the context of microblog computation. Since microblog contents are normally short and irregular, traditional senti- ment analysis algorithms cannot work well and meet the practical requirementsduo to its essential defects and ineffi- ciency. Microblogs topic opinion analysis is conducted by constrcuting a method of sentiment dictionary classification. Experiments on 32 000 sentences in 17 500 microblogs from 20 topics about collected from Tencent Weibo shows that the effectiveness of our proposed method.