基于关键句分析的微博情感倾向性研究  

Microblog sentiment orientation analysis based on key sentence analysis

作　　者：邵帅 刘学军 李斌 Shao Shuai;Liu Xuejun;Li Bin(School of Computer Science&Technology,Nanjing University of Technology,Nanjing 211816,China)

机构地区：[1]南京工业大学计算机科学与技术学院,南京211816

出　　处：《计算机应用研究》2018年第4期982-987,共6页Application Research of Computers

基　　金：国家自然科学基金资助项目(61203072);江苏省重点研发计划资助项目(BE2015697)

摘　　要：针对微博的情感倾向分析,提出了一种基于关键句分析的微博情感倾向性分析方法 SOAS(sentiment orientation analysis based on key sentence analysis),实现了从句子级到文档级的情感分析。首先,利用关键句抽取算法得到微博关键句,关键句抽取主要考虑位置属性、关键词属性和词频句子频特征这三类属性;之后,结合依存句法分析提出了影响情感倾向的七种词性搭配,并针对这七种搭配给出了六种情感计算规则,计算关键句的情感倾向值;最后,以关键句得分为权重,对所有关键句的情感倾向值加权求和得到微博的情感倾向。实现结果表明,基于关键句分析的微博情感倾向算法的情感分析,比同类算法的准确率高出了10.55%,提高了情感分析的准确率,具有高效性。This paper put forward a microblog sentiment orientation method SOAS based on the analysis of special sentences,it achieved sentiment analysis from sentence level to the document level.Firstly,it extracted the key sentences of microblog sentiment by key sentence extraction algorithm.This algorithm mainly considered 3 characteristics,namely location,keyword and the frequency characteristic of word and sentence frequency.Then,it computed the value of key sentences’sentiment orientation by 6 emotion calculation rules.These rules were given by 7 kinds of collocation of part of speech which would affect sentiment orientation.The 7 kinds of collocation of part of speech were extracted by dependency parsing.Finally,it used the value of key sentences as weights to weight sum all key sentences to judge sentiment orientation.The experiments demonstrate that the SOAS improves the accuracy of sentiment analysis.Compared to similar algorithms,its accuracy is 10.55%higher than that of others.It improves accuracy while having high efficiency.

关 键 词：情感分析 倾向性分析 关键句 依存句法分析 观点挖掘 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]