Research And Implementation Of Product Attribute And Inclination Recognition Technology From Comments Of E-Commerce And Weibo

Nowadays, the tide of electronic commerce and social network is sweeping across the world. Users usually express their opinions for the products after online shopping. These opinions are focused on the users as the center, reflect the users’ experiences, contain the users’ perceptions of product attribute, functions and performances. However, with the rapid growth of the number of online reviews and the complication of comments, it is difficult to get useful information manually according to the content of reviews. Therefore it is urgent to have some automatic approach to collect user comment information, analyze users" comments of product attribute and identify the users’ reviews. The opinion mining about the product attribute is based on the background and develops very rapidly.Mining product attribute, Ming the users’ opinions for the product attributes and recognizing the polarity of opinions are three key issues of opinion mining. This paper research the three key issues deeply, and concerning about the language characteristic of Chinese and the features of online comments, namely the comments usually don’t abide by the rules of grammar and sentences often omit the subjective. We pay much attention to analyzing the extraction problem of implicit subjective sentence, and identify the real subjective word after analyzing.The specific contents include:(1)The evaluation and assessment of feature words and opinion words. We use POSEM algorithm to extract the tuple of feature targets and opinion targets. Meanwhile, the syntactic structures of product reviews on the Internet are of free style, many sentences don’t have complementary sentence structures, so we propose an approach to extract implicit subjective target. It will enhance the recall and precision.(2) Determine the critic attitude, which means finding product attributes related to the opinion of the reviewer polarity from reviews. To find the point of view of polarity from user comments, first of all, we need to find evaluation vocabulary. After that, we make full use of adjective corpus, adverb corpus and domain corpus to judge the polarity evaluation vocabulary. Due to the fact that the polarity the word represents is often connected to the specific domain. The same word may have the opposite polarity in the different domains. (3) The design and implementation of the goods review information analysis tools, the tool includes the automatic extraction of comment information from Webpage’s DOM tree’s specific label, data preprocessing, subjective sentence extraction, the evaluation and assessment of word extraction, evaluation of word polarity judgment function.