摘要： 为研究情感词对情感倾向分析的作用,提高微博情感分析性能,提出一种情感词图模型的方法,利用PageRank 算法计算出情感词的褒贬权值,将其作为条件随机场模型的特征,预测具体语言环境下的情感词倾向。结合具体语境下的情感词倾向,利用支持向量机模型进行微博语料的主客观分类和情感倾向分类。实验结果表明,图模型构造的情感词典增加了具体语境下情感词倾向预测的准确性,具体语境下的情感词倾向预测对主客观 分类和情感倾向分类有明显的改善。

关键词: 图模型, 情感词, 条件随机场, 支持向量机, 网页排序算法, 倾向性分析

Abstract: For the further research of the function of emotional words on emotional analysis and the improvement of microblog emotional analysis method,this paper proposes a research approach to construct emotional words graph model using relations between emotional words. The emotional value of appraisal calculated by PageRank algorithm and trained as the feature of conditional random field model so as to forecast the tendency of emotional words in specific situations, through which subjectivity classification and emotional tendency analysis of microblog can be made when integrated with Support Vector Machine(SVM) model. Experimental results show that emotional lexicon constructed by graph model enhances accuracy of the prediction of emotional word in specific situations which is also helpful for subjectivity classification and emotional tendency analysis of mircroblog.

Key words: graph model, emotional words, condition random field, Support Vector Machine ( SVM ), PageRank algorithm, orientation analysis

中图分类号: 

• TP18