基于复杂句式的文本情感倾向性分析  

TEXT SENTIMENT POLARITY ANALYSIS BASED ON COMPLEX SENTENCES

作　　者：邸鹏[1] 段利国[1,2] 

机构地区：[1]太原理工大学计算机科学与技术学院,山西太原030024 [2]武汉大学软件工程国家重点实验室,湖北武汉430072

出　　处：《计算机应用与软件》2015年第11期57-61,共5页Computer Applications and Software

基　　金：国家重点实验室开放课题项目(SKLSE2012-09-30);山西省自然科学基金项目(2013011015-2);太原理工大学学科建设专项项目

摘　　要：在文本的情感倾向性研究中缺乏对多种复杂句式的有效分析,而复杂句式中多种情感共现的特点使得传统的情感分类器对复杂句式的情感分析效率不高,所以提出一种新的可以对复杂句式进行有效情感分析的情感分类模型。该模型充分分析了汉语中复杂句式的结构特点,通过已有资源构建中文情感词典、关联词表、否定词表,并提出了一种复杂句式模型来匹配各种复杂句式。最后将该复杂句模与朴素贝叶斯分类器相结合,得到新的针对复杂句式的情感分类模型。在实验中,新的情感分类模型在准确率、召回率、F值上都比传统的情感分类器有了明显的提高。实验证明该模型能更好的分析各种复杂句式的情感。There lacks the effective analysis on various complex sentences in the study of text sentiment polarity analysis, while traditional sentiment classifier has low sentiment analyse efficiency on complex sentences because of the characteristics of multiple emotions co- occurrence in complex sentences. Therefore, in this paper we propose a new sentiment classification model which can be used to analyse the sentiments of complex sentences effectively. The new model fully analyses the structural characteristic of Chinese complex text, and constructs the Chinese sentiments dictionary, associated words table and negative words table through existing resources. Moreover, it proposes a complex sentence model to match all the complex sentences. Finally we combine the new complex sentence model with the naive Bayes algorithm, and derive the new sentiment classification model for complex sentences. In experiment, the new sentiment classification model has an obvious improvement in accuracy rate, recall rate and F-measure than the traditional sentiment classifier. Experiment proves that the new model can analyse the emotion of various complex sentences better.

关 键 词：情感倾向性 复杂句式 情感分类模型 朴素贝叶斯 

分 类 号：TP391.1[自动化与计算机技术—计算机应用技术]