【作者】 吴苏红；

【导师】 王素格；

【作者基本信息】 山西大学 ， 控制工程， 2011， 硕士

【摘要】 随着人们生活水平的提高,旅游已成为人们生活的重要组成部分。与此同时,关于旅游景点的网络评论也越来越多。这些评论对于潜在型游客和各地景点管理商都是非常重要的信息资源。对于一般游客,在出游之前,可以通过网上评论了解其他游客对某景点的看法,规划自己的旅游行程。而对于景点管理商可以通过景点评论了解游客对景点的意见和态度,以便提高旅游景点的服务质量。但是,人工地逐篇阅读大量的旅游景点评论,需要花费许多的时问和精力,阅读者极有可能会“迷失”在其中,无法正确识别和利用其中有价值的观点信息。为了准确、高效地挖掘出游客感兴趣的观点信息,对文本进行情感倾向性分析是需要解决的关键问题之一本文利用词对间的依存关系,研究了评论文本的情感倾向分类和特征-观点对抽取方法。本文的主要研究工作如下：(1)基于规则的组块获取为了抽取对情感倾向分类有用的信息,本文利用了词对问的依存关系,构建了获取含情感倾向组块的规则。实验结果表明,基于规则的方法获取组块是可行的。(2)基于组块特征的评论文本情感倾向性分类对于旅游景点评论文本的情感倾向分类的特征选择问题,本文将利用获取的组块与情感词相结合作为情感倾向分类的特征。通过对旅游景点评论的情感倾向分类实验,结果表明,采用组块信息可以提高文本情感倾向分类的性能。(3)特征-观点对的抽取特征-观点对的抽取是观点挖掘中重要的研究课题之一,本文利用依存语法对句子的分析,研究了评论文本中特征-观点对的抽取方法。利用词对间的依存关系,先构建了获取含有评价对象和观点词语组块的规则以及候选评价对象的识别算法。在此基础上,设计了具有情感倾向的特征-观点对的抽取算法。通过实验验证了方法的有效性。更多还原

【Abstract】 With the improvement of people’s living standard, tourism has become an important part of people’s lives. Meanwhile, the online scenic spots reviews will be more and more. These reviews are considered as significant reference information for potential visitors and local scenic spots managements. Visitors have utilized this piece of this information to understand view of other visitors and plan trips through read online comments before traveling. In order to improve tourist attractions of service quality, managements of scenic spots may understand the opinions and attitudes about scenic spots. However, it needs to spend a lot of time and energy to artificial read mass reviews, and readers may have "lost", it is unable to identify and using the valuable information. In order to accurately and efficiently mine opinion information that is interested for visitors. Text sentiment orientation analysis is one of the key problems need to solve.This paper studies the review texts sentiment orientation classification and the method of extract the feature-opinion in review texts based on dependency relation. The major works of this thesis focuses on the following:(1) Getting chunks based on the rulesIn order to extract the useful information about sentiment orientation classification. By using the dependency relation between word and word words, this thesis constructs the rules to obtain chunks which contain sentiment orientation. Experimental results show that the method based on rule obtain chunks is feasible.(2) Review texts sentiment orientation classification based on chunk featuresThe thesis utilizes chunks combined with emotional words as features of sentiment orientation classification. Through the experiment of sentiment orientation classification about scenic spots reviews, experimental results show that adopting chunk information can improve the performance of text sentiment orientation classification.(3) Feature-opinion extractionFeature-Opinion Extraction is one of the key researches in the area of opinion mining. This thesis studies the method to extract the feature-opinion in review texts based on dependency grammar. By using the dependency relation between word and word, we construct the rules to obtain chunks which contain evaluation object and opinion word, as well as the algorithm to identify candidate evaluation object. On this basis, we design an algorithm to extract feature-opinion with sentiment orientation. Experimental results prove the method is effective.更多还原