Research On Extraction Method Of Implicit Evaluation Objects In Online User Reviews

With the rapid development of e-commerce in China,online shopping has entered people's daily life.Because of the asymmetry of information,it is difficult for consumers to understand the real value of the goods.Meanwhile,the online user reviews provide useful information for the user's purchase decision,so more and more scholars pay attention to the opinion mining of online comments.As an aspect of opinion mining,evaluation objects have been studied extensively.However,the research on the evaluation objects is mainly focused on the explicit evaluation objects,while there is only a few of research about implicit evaluation objects.The research on implicit evaluation objects not only can improve the classification accuracy of evaluation objects in research field,it can also help enterprises to find the hidden views of consumers in the user reviews.The consumers can get more real information and accurate comments by extracting implicit evaluation objects.Based on this,this paper studies the extraction method of implicit evaluation objects in user reviews.The content of the research is summarized as following:(1)Data preprocessing.By using the data capture tool,we get a lot of real comments from the site of Taobao.com.And then we using the word segmentation tool to split the sentence and words,while mark the part of speech,select the feature,express the text with vector representation,and so on.Because the spatial dimension of the initial text feature we got is very high,this paper uses the particle swarm optimization algorithm based on simulated annealing to extract the feature set to reduce the spatial dimension of the feature words.The experimental result shows that the spatial dimension of the feature words is reduced from 425 dimension to 296 dimension by using this method,and prove that this method can effectively select the feature.(2)Clustering analysis of explicit evaluation sentences.This paper divides the evaluation sentences into explicit evaluation sentences and implicit evaluation sentences,and carries on the text clustering research on the explicit evaluation sentences.Because the text vector space dimension is still high,this paper adopts the fuzzy c-means clustering algorithm which is suitable for high dimensional datasets.Aiming at the characteristics of fuzzy c-means,this paper proposes an improved fuzzy c-means algorithm based on simulated annealing.By controlling the iterative process of fuzzy c-means algorithm,the improved algorithm effectively avoids being trapped in local optimum.Through the experiment,the explicit evaluation sentences are clustered into nine categories and the category name is set for each category.The experimental result shows that the improved fuzzy c-means algorithm based on simulated annealing can reasonably classify the text.(3)Extraction research of implicit evaluation objects.After the text clustering of the explicit evaluation sentences,the evaluation sentences of same category are classified into a document set.In this paper,we use the association rule algorithm to mine the association rules between the categories,the evaluation objects and the evaluation words in different document sets.And then we use these rules to extract the implicit evaluation objects.At last,by comparing the experimental results,the accuracy rate of the implicit evaluation objects extraction method proposed in this paper is 75.26%,which can effectively improve the accuracy of text classification.