基于语义的微博短文本倾向性分析研究

摘要

通过结合情感词库和微博语义特征，采用向量空间模型的方法表示微博文本，提出一种融合模式匹配和机器学习的微博文本倾向性分析方法。对分词之后的微博文本，先提取出情感关键词，再匹配分析出的几种模式抽取情感评价短句、情感短句和微博表情符号等其他情感特征共同组成情感特征序列，最后采用多步分类的支持向量机方法得到微博文本的情感倾向。通过实验与其他评测结果进行对比，表明该方法是有效的。%Through a combining emotion thesaurus and semantic features of micro-blog,using space vector model to represent microblogging text,this paper proposed a method of microblogging text orientation analysis based on pattern matching and ma-chine learning.For the text after participle,it firstly extracted the sentiment key words,and then matched the key words and analyze several models to extract text phrases of sentiment evaluation,emotional phrases,micro-blog emoticons and other emo-tio-nal features characteristic constitute micro-blog sentiment characteristic sequence.Finally,it made use of SVM machine learning algorithms to get sentiment tendencies of micro-blog text.Evaluation results are compared with other paper,the experi-mental results show that this method is effective.