Research On Attribute-level Sentiment Analysis Method Of Case Microblog Comment

At present,hot-spot cases have attracted more attention from people,and a lot of discussions have been conducted on social media platforms such as Weibo,which aroused public opinion about the cases involved,and could easily have a negative impact on the court’s trial work.Identifying the aspect words in the Weibo comments involved in the case,mining the opinions of the aspect words and judging their emotional polarity can help the court to monitor the public opinion involved in the case and evaluate the trend of public opinion.Aspect-level sentiment analysis for case microblog comments has problems such as lack of case aspect-level sentiment analysis corpus,lack of effective use of case domain knowledge in aspect word extraction,and insufficient utilization of the relationship between opinion words and aspect words.This paper focuses on the tasks of aspect word extraction and aspect-level opinion mining for microblog comments involved in the case,and has completed the following tasks.(1)Due to the lack of publicly available aspect-level sentiment analysis corpus of the involved Weibo,this paper constructs an aspect-level sentiment analysis dataset of the involved Weibo comments.First,a large amount of Weibo comment data involved in the case was crawled from Weibo pages through crawling technology.Secondly,formulate a labeling system to manually label the aspect words,opinion words and emotional tendencies of the microblog comments involved in the case,and construct an aspect-level sentiment analysis corpus of the microblogs involved in the case to provide data support for subsequent research.(2)The extraction of microblog aspect words involved in the case aims to identify the case aspect terms evaluated by users from the microblog comments.The existing methods do not consider the domain characteristics of the microblog involved in the case,that is,the comments usually revolve around the case keywords appearing in the text.open discussion.Therefore,this paper proposes a method for extracting aspect words of involved microblogs based on keyword structure coding.The method comprehensively utilizes the case keyword information of the microblog body through the structure coding mechanism,and integrates it into the comment sentence representation through the interactive attention mechanism to guide the extraction of aspect words.Experimental results demonstrate the effectiveness of the method in the extraction of related microblog aspect words.(3)The aspect-level opinion mining task of the microblog involved in the case aims to extract the opinion words corresponding to the aspect words and judge their sentiment tendency,including the two sub-tasks of opinion word extraction and aspectlevel sentiment classification.The difficulty of this task lies in how to model the relationship between aspect words and opinion words,so as to use the extracted opinion words to predict sentiment polarity.Considering that there is a certain syntactic relationship between aspect words and opinion words,at the same time,opinion word extraction and aspect word sentiment analysis are two mutually promoting tasks.Therefore,this paper proposes a case microblog aspect level based on graph attention network.Opinion mining methods.The method utilizes syntactic relations and relationaware graph attention network to jointly train opinion word extraction and aspect-level sentiment classification.It can improve the precision of opinion mining.Experimental results demonstrate the effectiveness of the method in aspect-level opinion mining of implicated microblogs.(4)This paper designs an aspect-level sentiment analysis prototype system for case microblog comments.The system can realize the aspect word extraction function of the microblog comments involved in the case through the aspect word extraction model based on keyword structure coding proposed in this paper,and realize the opinion words of the microblog comments involved in the case through the opinion mining model based on the graph attention network proposed in this paper.Extraction and aspect-level sentiment classification functions.