Research On Product Reviews Sentiment Classification Based On Platform Of Microblog

The emergence of Web 2.0 had been promoting the development of e-commerce and Microblog rapidly. As a popular Web 2.0 Internet application, Microblog has been produced a mass of product reviews text which allows users to express their views that contains emotional information. Product reviews on Microblog platform is a kind of customers’ emotional expression whether the products is good or not. The benefits of studying this information by sentiment classification has two aspects: not only can help customers to make a good decisions according to reviews, but also help producers get the reviews message in time which to be reference to improve their product quality and adjust appropriate marketing strategies to make their product popular.As a new model of product reviews in Internet, product reviews on Microblog platform combines the features of Microblog text that wide range topics, colloquial expression and language fragmentation, etc. Compared with the traditional text, sentiment analysis for product reviews on Microblog platform will encounter more challenges and difficulties. At present, the research object of product reviews sentiment analysis is product reviews on traditional e-commerce or product BBS. Former domestic sentiment analysis research of product reviews on Microblog platform is still at an early stage. Therefore, the research of Microblog product reviews sentiment analysis has a highly theoretical and practical value.Product reviews on Sina Microblog platform as objects of the study, sentiment analysis method which combined sentiment lexicon and semantic rule was introduced. Users’ views will be analyzed base on the sentiment results which combine theory study and application as follow.First, according the feature of Microblog, a sentiment classification method on the Microblog product reviews was proposed. This approach contains four aspects: first of all, according to colloquial expression of Microblog, new word set was extracted by NLPIR system. And then dig up the new emotional words set and new evaluation object set from the new word set. Secondly, a sentiment lexicon based on the Hownet was built. In order to improve the recognition rate of emotional words, the new emotional words set was combined with lexicon. Thirdly, we proposed a recognition rule for evaluate collocation based on the analysis of the part of speech of evaluation objects. The recognition rule of evaluate collocation is centered on emotional words, order by the importance of the part of speech of evaluation objects and measure by the words distance. Finally, considering the impact of degree adverbs and negative adverbs for text rhetoric, we construct six kinds of combination of emotion, the values of reviews text emotional polarity will be calculated. Experimental results show that this method can effectively extract the collocation and classify sentiment of Microblog.Secondly, we quantified the users’ views by above method. Considering the forwarding numbers and review number of Microblog, analysis indicators were built to analyze the users’ views: firstly, users’ views analysis based on the recognition of product attribute. We show users’ views about evaluation of product attribute through chart, and analyze the advantage and disadvantage of product attribute to propose the improvement direction. Secondly, Analysis of product evaluation based on Analytic Hierarchy Process. Factors about product evaluation will be sorted to calculate the weight by Analytic Hierarchy Process. And combined with the product attribute recognition, the reputation of products will be measured.