Design And Implementation Of Product Review Sentiment Analysis System

Today,Online shopping has become the main mode of consumption.The proportion of online shopping consumption increased year by year in China.More than half of consumers will publish comments about the product when they shopping,and more than seven layers of consumers will check the product review when they purchase of goods.So,the product review has become an important part of the consumer purchase decision.According to the SINA News,Tmall 11 global Carnival total transaction is over 120 billion 700 million,covering more than 235 countries in2016.And the product review has come to the TB level.For such a large amount of data,consumers scan product reviews one bye one have become unrealistic.But if you read only part of the product review will often causing misunderstand.When user purchasing of products,they query product parameters in various types of Web site.Sometimes a single Web site can not satisfy these users,they need to view product parameters by search lots of professional web.In order to solve this problem,it is necessary to have a strong analysis of product reviews tool,so that consumers can be comprehensive and timely get product review information.And the tool need to help consumers make purchase decisions,reduce the complexity of online shopping.The main contents of this paper are as follows.(1)Acquire and preprocessing of product review.Analysis the technology of Jingdong's anti web crawler,Taobao's anti web crawler and Amazon's anti web crawler.For different anti web crawler,propose the solution of anti web crawler,and use web crawler to get product review.Analysis product review data.For each problem in product reivews,we propose ways to deal with problem.And hope that these work will help the next step in emotional analysis,reduce the amount of manual work.(2)Emotional Analysis of Product Reviews.According to the results of the linguistic analysis of product reviews,use the Bootstrapping algorithm to combine the linguistic rules and PMI method to construct the opinion lexicon and target lexicon.Through the opinion lexicon and target lexicon,the use of rules to extract target word-opinion word.And establishing mapping relationship of product feature and target word to convenient to show the analysis with the results of the work.(3)Product reviews emotional analysis results show.Using the Django framework,design and implement product reviews sentiment analysis system.The system has the product parameter information module,the brand parameter information module,the product emotion information module,the brand emotion information module,the contrast module and the recommendation module.Parameter information module in order to reduce the tedious steps when user querying product information.Emotional information module from different angles on the emotional analysis results,so that users can easily get the product reviews contained in the emotional information.Contrast module and recommended module is a combination of product reviews emotional analysis results and parameter information to help users buy goods.And this paper introduces point mutual information technology,syntax analysis,semantic technology platform and Bootstrapping algorithm.Then,this paper introduces how to crawl the product reviews and how to deal with the product reviews.According to the product reviews,building an opinion lexicon and target lexicon.And the extract target word-opinion word,establish mappings between product characteristics and target word.Finally,we can recommend product what users needs.This paper introduces a method of acquiring and preprocessing product review.And show the way to build an opinion lexicon and target lexicon,extract target wordopinion word,establish mappings between product characteristics and target word.This system can help users to make purchase decisions,reduce the complexity in shopping.