The Research On Technologies Of Product Review Oriented Opinion Mining

The diversity and convenience of the online shopping mak more people tend to choose shoping on the Internet.But the number of product reviews is large and unstructured and inform,the exits of these problems bring some difficult to opinion mining.How to automatly extract product features and opinon words from product reviews is a host issue for research.In this article,we aim to research fine-grained opinion mining.Firstly,the method of associate relation to mining product features and opinon words is lack semantic of words.To solve this problem,we propose to join dependency relation to extract product features and opinon words.The dependency relation can generate the relaton type of product features and opinon words.So we can use this to extract product features and opinon words and get the score of product features and opinon words.next to use the score of associate relation.Then we can get the final score of product features and opinon words.Through this,we can improve precision of the result of extracting.Compare with other methods,we improve the result of extracting 5%,3%,and get the pair of product features and opinon words,this can help us to summarize customer reviews.Secondly,we come up with a method of emotion calculation about joint append comments to deal with the problems which the same user changes the original comment about goods and so on.We analysis the relationship of original and additional comments and decide whether comments of the same user have the attribute of new goods or the difference of emotional tendency of the attribute of same good.The results of experiments show the method proposed in the paper of affective forecasting the sentiment score.The paper regards the comments of goods as the research object.This paper propose two methods which respectively are the detection method based on the attribute of goods and opinion words and the method of emotional calculation combined with additional comments.In this paper,the research results help extract product features and opinion words,and know disadvantages and disadvantages of every arrtibute of goods and help users decide to buy what kinds of goods.