Research On Opinion Mining With Consideration Of Implicit Product Features

With the development of web and the Business to Customer model, the retailers start to usenetwork to sell their goods, and we find that there are massive reviews in Wechat, blogs andE-commerce. Knowing customers’ opinions for the products or service, the retailers can improvetheir products or service to meet customers’ needs; therefore the retailers are paying more andmore attention to the social network comments. However if the retailers get useful informationfrom these massive reviews through manual work, it will be very time-consuming. Therefore, weneed opinion mining to mining the useful information in the reviews.This paper considered Chinese comments on the network as the research object.Text analysiscan be divided into three parts: automated building of product features’corpus and opinion words’corpus, the extraction of opinion target association and analysis of reviews’ emotional tendency.This paper mainly consisted of three parts, which are as follows.(1) On the basis of the word segmentation results, this paper used the phrase rules to identifyproduct features and opinion words. The experiment results proved that this method can improvethe accuracy of the word segmentation operation. And this paper put forward a method that canautomated build product features’ corpus and opinion words’ corpus. This method uses therelationship between product features and opinion words to build the corpus, the experimentproved that this method can improve the accuracy of the corpus building.(2) Through studying Chinese reviews on the Internet, we found that these reviews on thewebsite were hardly standard in grammar, and even having many syntax errors. Therefore, thispaper used opinion to identify the opinion target association rather than used syntax. There aredefects on vague opinions in existing implicit products feature extraction methods; therefore, thispaper proposed a novel context-based implicit feature extracting method. Experiments proved thatthe presented method in feature extraction of implicit products has better performance onaccuracy.(3) In the emotional tendency analysis process, this paper used the emotional polaritydictionary and the adverb in the reviews to analyze reviews emotional tendencies. Experimentsproved that this approach can lead to emotional tendency analysis results with higher accuracy.(4) This paper designed a generic reviews opinion mining system, and building corpus,extracting the opinion target association and emotional tendentious analysis and its final resultsare realized among that system.