导　　师： 朱征宇

学科专业： 081202

授予学位： 硕士

作　　者： ;

机构地区： 重庆大学

摘　　要： 随着Web技术的发展,网络上现存的主观意见表述文本陡增。主观WEB文本不仅能给其他用户提供客观的参考价值,并且产品生产方也可以及时根据用户的意见言论来调正自己的产品设计以及营销策略。因此主观意见文本的处理也随之受到人们的重视。近年来也成为了WEB数据挖掘的热点。 在中文意见挖掘中,相关学者都将如何准确得到意见词的语义情感倾向作为他们的着手点,对准确获取特征以及特征之间的关系相对研究较少。而特征的相关层次关系分析将对后面的语义情感分析和对整篇评论的理解起到重要作用。因此,为了研究特征词的抽取方法和特征层次关系的获取方法,本文主要做了如下几方面的研究工作: ①由于语料在中文文本挖掘中的重要作用,描述了不同语料在意见挖掘中的运用,分析了不同语料对于特征词提取的作用。并针对不同语料介绍了不同的特征词抽取算法。 ②描述了中文语料中能够体现概念上下位与部分整体关系的特定语法结构,并从统计的角度探索和深入地分析了汉语“的”字结构短语中的概念上下位关系,总结出一种利用Tdf//idf公式判定“的”字前后概念上下位关系的方法。 ③研究给出了基于网络的PMI统计算法的一种新应用。通过使用特定语法结构短语在现有比较成熟的网络搜索引擎中进行搜索,来获取词汇和加入特定语法结构的短语在网络中被搜索到的次数,用于判断词汇之间的关系。 ④分析了现有中文评论特征词与意见词所形成的词对的抽取算法。并将Bootstrapping算法和关联规则算法对特征词的抽取效果作了对比和分析。 通过较充分的实验,验证了PMI算法用于特征层次提取的可行性和有效性。论文研究为基于网络的PMI算法在中文意见挖掘的应用提供了新的可供借鉴的方向和思路,具有较� With the development of Web technology, the subjective text in the network increased sharply. Subjective web page not only gives other users objective reference,but also producer can adjust his product designing and marketing strategy according to opinion from users. Therefore, subject text is catching more people’s eyes .In recent years, Data mining has become a hot spot. In Chinese opinion mining area,relative researcher pay more attention to get the semantic orientation accurately,on the another side of the coin,less people do researching about archiving the Feature hierarchy accurately。However,feature hierarchy can play very important part in semantic orientation analysis and opinion paper understanding. Its contribution to the main content of this article: ①Because of the important function in Chinese text mining, describing the application of different corpus in opinion mining and analyzing different function of corpus for feature extracting. For different corpus, the different extracting algorithm of feature hierarchy was obtained. ②Introducing some special phrase syntaxes which contain‘is-a’and‘part of’relation, and trying to analyzing‘is-a’relation of concepts in the de phrase. At last, summarizing a method which using tdf//idf function to get the‘is-a’relation in the de phrase. ③Getting a new application of PMI statistic algorithm after researching. Using special phrase syntax as key word to query sophisticate business search engine, and then using the number queried by searching engine to count the words’hierarchy relation. ④Analyzing the feature-opinion word pair extracting algorithm in Chinese Opinion Mining. The contrast between Bootstrapping algorithm and Association Rule Mining algorithm was made. Executable of PMI algorithm’s application in extraction of feature hierarchy has been proved by the experiment. And researching in this paper provides a new thinking and orientation for PMI algorithm application in Chinese Opinion

关 键 词： 意见挖掘

领　　域： [自动化与计算机技术] [自动化与计算机技术]