Aspect Level Opinion Mining And Application Research Based On E-commerce Reviews

With the rapid development of network technology and the popularization of connected devices such as personal computers,smartphones and tablets,people’s shopping methods are gradually shifting from offline to online.The booming e-commerce platform is serving thousands of households.When potential consumers choose products on the e-commerce platform,in addition to paying attention to the information of the products themselves,they also obtain evaluation information of other consumers from product reviews,which is more true,direct,and effective.In order to help businesses get market feedback more quickly and guide potential consumers,opinion mining technology is used to automatically identify and organize comments on e-commerce reviews.If only coarse-grained sentiment analysis of the overall e-commerce reviews does not provide effective information,it is necessary to dig out multiple aspects of the products in the reviews and the emotional tendencies that consumers hold in these aspects.In response to these needs,this article proposes an aspect-level opinion mining model OMTS(opinion mining method based on topic model and SVM),which uses a combination of topic model and classification methods to obtain consumers’ opinions on various attributes of products in e-commerce reviews.Information,and analyze the sentiment tendencies.The main research work of this article is as follows:(1)Crawl 4000 notebook e-commerce reviews and 5000 mobile phone e-commerce reviews on the Jingdong website,mark the reviews as negative and positive categories according to the scores of the reviews,and pre-process comments.(2)The topic model is used to identify aspects in comments.In order to distinguish between opinion words and sentiment words in the review text,this article integrates the three existing Chinese sentiment dictionaries and uses dictionary-based matching methods to remove sentiment words in the text and remove the remaining words in the text.The text uses topic models for aspect recognition,thereby improving the accuracy of aspect recognition.The topic model can be used to obtain multiple aspects of the corpus as a whole,and each comment belongs to a certain aspect with the highest probability.This article classifies the comments according to the aspects,and evaluates the results of aspect recognition by constructing an industry knowledge graph.Conduct sentiment analysis and opinion content mining for the comments under each aspect category.Sentiment analysis uses a classification algorithm based on machine learning,uses the N-gram model to map comments into word vectors,trains and tests support vector machine classifiers;filters verbs,adjectives and adverbs that contain sentiment polarity in the comments,and this article still uses The topic model is used to mine opinion content.(3)In view of the crawling data set,we use the method proposed in this paper to carry out the experiment of view mining.Through the comparative experiment,we prove that the aspect level view mining proposed in this paper has a good result.At the same time,this article also designs and implements a viewpoint mining system,to show the users the aspects identified in the comments,the content of each aspect,and the emotional tendencies.