基于酒店中文评论情感倾向分析

发布时间：2019-01-03 14:23

【摘要】：目前随着互联网的发展,特别是在进入Web2.0时代,互联网上产生了大量的信息。其中这些信息大部分都是由互联网用户创造的,而这些信息很大一部分是以文字的形式出现在互联网上。人们会使用电子商务网站进行购物,会留下对产品的评论;人们会通过新闻网站阅读最新的新闻消息,留下自己对该新闻的看法与观点;人们也会在各种社交网络平台发表自己的各种情感。而对于这些文字信息的挖掘产生了自然语言处理领域的各种技术与研究方向,其中对文本情感的分析是目前热门的研究课题。通过对用户发表的文字信息使用自动的情感分析技术,在现实中有许多应用场景,例如产品推荐系统、网络舆情分析系统、政务决策系统等。中国作为拥有世界互联网用户数量最多的国家之一,对于中文文本的情感分析需求日益增加。本文对酒店领域中文评论情感倾向分析应用进行了研究,主要任务有:(1)构建了一个酒店领域的特征词典和情感词典。情感词典构建中使用常用的中文情感词典作为基础词典,抽取酒店领域评论中的情感词,并且使用词向量计算词之间的相似度的方法来确定候选情感词的情感倾向,以此来扩展词典。(2)利用句法依存关系分析,特征词典与情感词典,对酒店评论进行细粒度情感倾向分析。通过抽取特征词,情感词与修饰词等组成的情感搭配元素组合并进行情感倾向计算,获得酒店在各个属性特征上的情感倾向评价。(3)对酒店整条评论进行情感倾向分析,采用情感词典与机器学习结合的方法。首先对评论使用情感词典计算情感倾向值,挑选出每一类中情感倾向值明显的评论作为训练集,并且使用多特征融合方法结合多个SVM-KNN分类器能力进行分类。实验表明优化挑选训练集能提高分类器效果,多特征融合方法比单一分类器效果略有优势。(4)设计实现了一个基于评论情感分析的酒店检索系统。目前的酒店预订网站对酒店的检索都是使用酒店的客观指标,比如酒店位置,价格范围等。而本文的系统不仅使用酒店的客观指标并且通过对酒店评论进行情感倾向分析,得到用户对酒店的主观评价指标,比如性价比是否高,环境是否舒适等。结合用户对酒店的客观与主观指标要求,进行酒店检索,将结果返回给用户。
[Abstract]:At present, with the development of the Internet, especially in the Web2.0 era, the Internet has produced a lot of information. Most of this information is created by Internet users, and a large part of this information appears on the Internet in the form of text. People will use e-commerce sites for shopping, will leave comments on the products, people will read the latest news through the news site, leaving their own views and views on the news. People also express their emotions on various social networking platforms. The mining of these text information has produced a variety of technologies and research directions in the field of natural language processing, among which the analysis of text emotion is a hot research topic. Through the use of automatic emotion analysis technology for the text information published by users, there are many application scenarios in reality, such as product recommendation system, network public opinion analysis system, government decision system and so on. As one of the countries with the largest number of Internet users in the world, China has a growing demand for emotional analysis of Chinese texts. The main tasks of this paper are as follows: (1) A characteristic dictionary and an emotional dictionary are constructed. In the construction of emotion dictionary, the commonly used Chinese emotion dictionary is used as the basic dictionary to extract the emotion words from the hotel domain comments, and the method of calculating the similarity between the words by word vector is used to determine the emotion tendency of the candidate emotion words. (2) using syntactic dependency analysis, feature dictionary and emotion dictionary to analyze the fine grained emotional tendency of hotel comment. By extracting feature words, affective words and modifiers of emotional collocation elements and calculating emotional tendency, the evaluation of hotel's emotional tendency on each attribute feature is obtained. (3) the emotional tendency of the whole comment is analyzed. The method of combining emotion dictionary with machine learning is adopted. Firstly, the emotion dictionary is used to calculate the affective tendency, and the comments with obvious affective tendency in each category are selected as the training set, and the multi-feature fusion method combined with the ability of multiple SVM-KNN classifiers is used to classify the comments. Experiments show that the optimal selection training set can improve the classifier effect, and the multi-feature fusion method has some advantages over a single classifier. (4) A hotel retrieval system based on comment and emotion analysis is designed and implemented. At present, the hotel reservation website searches the hotel using the hotel objective index, such as hotel location, price range and so on. The system of this paper not only uses the objective index of the hotel, but also gets the subjective evaluation index of the hotel by analyzing the emotional tendency of the hotel comment, such as whether the ratio of performance to price is high, whether the environment is comfortable, and so on. According to the objective and subjective requirements of the users, the hotel retrieval is carried out and the results are returned to the users.
【学位授予单位】：华南理工大学
【学位级别】：硕士
【学位授予年份】：2016
【分类号】：TP391.1;TP181

【相似文献】