Fine-grained Chinese Text Emotion Analysis Based On Deep Learning And Rule Base Learning

At present,with the rapid development of e-commerce platforms and social platforms,a large number of netizens' opinions about social events and commentary data on commodities have been precipitated on the internet.Analysis of public comment can help public authorities monitor the social public opinion trends.Analysis of product reviews can sum up the user view of the product,help businesses understand user needs,and improve the product.Therefore,a large number of comments on the internet have high social value and economic value.We must rely on automatic emotion analysis technology to analyze the emotional tendency of users in comments from a large amount of review corpus.This article focuses on the characteristics of e-commerce platform product reviews and research emotional polarity analysis related technologies.This paper mainly completed the following work.Aiming at the inaccurate emotion analysis caused by the semantic transition in the product reviews,this paper designs a fine-grained emotional polarity analysis framework for different evaluation objects of the commodity itself,improves the accuracy of the comment analysis and eliminates the ambiguity.In addition,in the fine-grained sentiment analysis model,the CNN framework is introduced to improve the influence of unregistered words in the RNN framework on sentiment analysis with an accuracy of 95%.Aiming at supervised learning method relies on label data and number of products is large,this paper proposes a rule base learning based on deep learning.To a certain extent,it can improve the generalization.The framework first extracts the word part-of-speech,degree adverbs,polar words,thesaurus,negative words and other information in the commentary,and then models and learns the rule base of emotional polarity analysis.Emotional learning with a single type of product review lets you migrate to other types of product reviews.Finally,the paper carried out the visualization of the algorithm results.After fine-grained analysis of the comments,the results are visualized according to the evaluation objects,and the visualization of the comments focuses on the negative comments of the users.