Research On Sentiment Analysis Method Based On Product Comments

The network technology is under increasingly development and the public's faith in the network security,more and more people are using the network information to have interaction with each other.Network influence people's life,One of them is online shopping.The benefit brings by the network to the public is numerous,one of which is online shopping.Comparing to physical stores,the online goods have more option and comparability,and people are more likely to buy cheap goods.However,the online goods only have some parameters of objective existence,pictures,etc,customers can only judging by the information inside the product whether the goods online meet their own needs,lacking of touch and trying on experience.Most often,there is a big gap between the buying goods and the consumer's expectations.As a result,there is a series of changing or refunding problems.If it is possible to get some experienced information from the former goods' buyers-goods comments,that would be a great guiding significance for the consumer to buy right things.Due to the increase in the demand for comments,The number of comments on the rise.it is difficult to discover the helpful information from such a large scale.If we carried out in accordance with the comments of emotional tendency by category classification,it would be the same emotional tendency of comment in one category,so that consumers can according to their own needs to verify the categories to find useful information for themselves.And the research's emphasis of this article is to review the classification method.The research emphasis of this paper is an analysis of the emotional goods comments,judge goods comments emotional tendencies.This paper mainly research the sentiment analysis from follow two aspects:(1)Sentiment analysis based on the emotional dictionary.The mainly problem of this method is the expansion of the emotional dictionary.In this paper,on the foundation of referencing the previous technology,putting forward two emotional lexicon extension methods.One of the expansion of the emotional dictionary is based on the combination of syntactic structure template and PMI values.The other is the expansion of Chinese dictionary based on the English emotional dictionary.(2)Sentiment analysis based on the technology of text categorization,the difficulty is the choice of feature vector of the method.How to delete the useless redundant feature and keep the useful feature without changing the classification result,is the mainly studying problem in this paper.In the thesis,in this thesis and the method of feature selection is improved basing on the information gain.Putting the frequency of words and words of emotional intensity into the feature selection method of information gain.Considering the words under the emotional intensity and frequency making the feature selection.Finally,this thesis make the sentiment analysis applying these three methods to the goods comments,and through a series of contrast experiment to testify the effectiveness of the proposed methods.