基于平滑SO-PMI算法的微博情感词典 构建方法研究

对现有情感词典在微博情感分类中的适用性进行了分析,针对现有情感词典在微博中情感词覆盖度低的问题,整合现有情感词典资源,构建了一个微博基础情感词典,同时提出了一种基于拉普拉斯平滑的SO-PMI算法对微博基础情感词典中没有收录的情感词倾向性进行判断,最后利用微博情感词典与拉普拉斯平滑的SO-PMI算法对微博情感词典进行了构建,并对所构建微博情感词典的分类性能进行了实验。实验结果表明,该方法所构建的情感词典在微博情感分类中能达到较好的分类效果。

Research on Construction of Microblog Sentiment Lexicon Based on the Smooth SO-PMI Algorithm

Analyzed the applicability of the existing sentiment lexicon in the microblog sentiment classification. In view of low coverage of the existing sentiment lexicon, built a basic microblog sentiment lexicon by integrating the existing sentiment lexicon, and put forward a Laplacian-based smooth SO-PMI algorithm to judge emotional orientation of the words which not included in the basic sentiment lexicon, finally applied the microblog sentiment lexicon and the Laplacian smooth SO-PMI algorithm to construct the microblog sentiment lexicon, and tested the constructed lexicon classification capabilities. Experimental results showed that the constructed microblog sentiment lexicon achieved good effect inmicroblog sentiment classification.