文本情感分析综述  

Survey of text sentiment analysis

作　　者：杨立公[1] 朱俭[2] 汤世平[1] 

机构地区：[1]北京理工大学计算机学院,北京100081 [2]中国青年政治学院计算机教学及应用中心,北京100089

出　　处：《计算机应用》2013年第6期1574-1578,1607,共6页journal of Computer Applications

基　　金：北京市自然科学基金资助项目(4123094)

摘　　要：以文本颗粒度为视角,从情感词抽取、语料库和情感词典构建、评价对象与意见持有者分析、篇章级情感分析、实际应用五个方面对文本情感分析文献进行了梳理,并做出必要评述。指出当前情感分析系统的准确率普遍不高,进一步研究的重点在于:自然语言处理的研究成果在文本情感倾向分析中更广泛和贴切的应用;选取文本情感倾向分类的特征和方法;利用现有语言工具和相关资源,规范、快速地构造语言工具和相关资源并应用。This survey summarized the studies on text sentiment analysis in the view of granularity from the following five aspects： sentiment word extraction, sentiment corpus and dictionary construction, entity and opinion holders analysis, document level sentiment analysis, and text sentiment analysis applications. It pointed out that the current sentiment analysis system cannot gain high precision. Further research should focus on： widely and appropriately applying study achievement of natural language processing to text sentiment analysis; finding and choosing suitable features and algorithms in text sentiment classifications; utilizing the existing language tools and relevant resources in fast building standard language tools and resources and applying them.

关 键 词：文本情感分析 情感词 语料库 情感词典 意见持有者 

分 类 号：TP391.1[自动化与计算机技术—计算机应用技术]