Research And Implenmentation On Aspect-level Sentiment Analysis System Of User Reviews

As the Internet enters the Web 2.0 era,user-led content Internet product models have become popular.The Internet has become more closely connected with people’s lives.In the complex and diverse information,text still occupies the main part.For new users and merchants,sentiment analysis of existing user reviews is very important,which not only helps new users make purchase decisions,but also guides merchants to improve their products.Aspect level sentiment analysis is a sub-task of sentiment analysis,which aims to extract the product attributes mentioned in the text and judge the corresponding sentiment polarity.This research theme is of great significance for solving problems in practical applications.This article designs and implements an aspect-level sentiment analysis system for user reviews.The users can be merchants on the network platform or consumers on the Internet.The system can help users obtain detailed and accurate information to make the most efficient purchase decisions.It can also help businesses analyze the advantages and disadvantages of their products in a timely manner,understand the product attributes that consumers care about,and improve product quality in a targeted manner.In order to solve the actual technical pain points in the application of Internet public opinion analysis,the work of this article is carried out from the following aspects:1.In order to obtain the aspect category(implicit attribute)in the user review text,this paper proposes an aspect category detection method based on the fusion of recurrent neural network and external features.2.In order to solve the problem of error propagation caused by the two-stage method in practical applications and the mismatch of aspect categories and sentiment polarity,we propose a novel model(TG-GRU)based on the combination of GCN and Bi-GRU for detecting aspect categories and corresponding sentiment polarities simultaneously.The method has improved effect compared with the existing joint model.3.In order to better serve the target users of the system and apply the aspect-level sentiment analysis model to practice,this article implements a public opinion analysis system that integrates aspect category detection,aspect category sentiment classification,and joint modeling analysis.And we evaluated its effect.