基于语义特征的微博情感分析研究  

Study of microblog sentiment analysis based on semantic feature

作　　者：宁慧[1] 杨松[1] 赵勇[1] 徐丽[1] 

机构地区：[1]哈尔滨工程大学计算机科学与技术学院,黑龙江哈尔滨150001

出　　处：《应用科技》2016年第3期70-74,共5页Applied Science and Technology

基　　金：国家自然科学基金项目(61201084)

摘　　要：针对在微博情感分析中没有有效地构造情感词典的方法问题,提出基于《同义词词林》和微博检索系统的情感词典的构造方法。利用《同义词词林》对4部基础情感词典进行扩展形成一部情感词典,利用点互信息(PMI)公式计算情感词语的情感倾向值。针对情感分析中特征选择的问题,利用不同的特征组合通过对比实验选取了有效的特征组合。实验证明,构造的情感词典和选择的特征组合有效,能够明显提高微博情感分析的质量。For there is no effective method to construct the sentiment dictionary in the microblog sentiment analysis,this paper proposed a method for constructing the sentiment dictionary based on the Synonym Thesaurus and microb-log index system. The method uses the Synonym Thesaurus to extend the four basic sentiment dictionaries, forminga sentiment dictionary, and uses the pointwise mutual information（PMI） formula to calculate the sentiment tenden-cy value of sentiment words. For the problem of feature selection in sentiment analysis, the effective feature combi-nation is selected by comparing different feature combinations. Experiments show that the structure of the sentimentdictionary and the selection of the feature combination are effective, significantly improving the quality of microblogsentiment analysis.

关 键 词：微博 情感分析 情感词典 微博检索系统 语义特征 PMI公式 情感倾向值 特征选择 

分 类 号：TP391.1[自动化与计算机技术—计算机应用技术]