Design And Implementation Of User Comment Analysis System Based On ACSA

With the popularization and development of Internet technology,online shopping has been gradually integrated into people’s daily life.Especially in the environment of novel coronavirus epidemic,more and more people choose online shopping.Followed by a large number of user comments data,including users’ evaluation of goods and services,the sentimental orientation of these comments is very important to improve the quality of products and services.Therefore,it is of great significance to mine the user’s sentimental orientation in the comments.As a hot task in natural language processing,text sentiment analysis has important application value in daily life.At present,there is not a special platform for text sentiment analysis in the market.although some large companies may open the API interface to sell the model of sentiment analysis,the threshold of application in this way is high and it is difficult to apply to individual industrial and commercial households who do not understand technology.In addition,the API interface of text sentiment analysis opened by some large companies is only suitable for general text sentiment analysis,and it is difficult to meet the needs of users with more fine-grained sentiment analysis.Therefore,the development of a fine-grained text analysis platform based on user comment data has important application value.This paper uses fine-grained sentiment analysis task to mine the sentimental orientation of user comments,including users’ positive and negative feelings about specific goods and services,and users can observe the analysis results of more fine-grained goods.In addition,the whole system uses graphical interface operation to make the user operation more simple and convenient.According to the different degree of analysis of text data,text sentiment analysis can be divided into sentiment analysis for text data,sentiment analysis for sentences and aspect-based sentiment analysis.The aspect-based sentiment analysis can be divided into aspect term sentiment analysis(Aspect Term Sentiment Analysis,ATSA)and aspect category sentiment analysis(Aspect Category Sentiment Analysis,ACSA).However,the strategy adopted by existing ACSA models is to train individual parameters for each aspect category.Therefore,when there is the problem of insufficient category data in some aspects,it will lead to the underfitting of model training.In order to solve this problem,this paper proposes a joint model with shared sentiment prediction layer,and uses it as the sentiment analysis model of user comment analysis system.in order to accurately analyze the role of different sentiments expressed by users towards different kinds of goods.The main contents of this paper are as follows:1.The construction of user comment system.The system uses PyQt5 as the front-end development framework of the comment management system.The system includes the following functions:1.User login registration module,which provides checking and checking of account name and account password,account registration and account login function.2.Model training module,which provides visual parameter configuration,model training and model incremental training functions of the model.3.Data analysis module,the main functions include data input function,model data analysis and data visualization analysis function.2.The development of text sentiment analysis system.In this paper,a joint model of aspect category sentiment analysis with shared emotion prediction layer is proposed and implemented.The model is mainly composed of two sub-tasks,one is to detect aspect categories,and the other is to analyze the sentiment of aspect categories detected.The model shows its excellent performance under several public data sets,and the corresponding results are presented at the CCL2020 conference.3.Visualization partThe data visualization part uses the drawing toolkit in the PyQt5 framework to draw a bar chart for the data analyzed by the sentiment analysis model.4.Test partWe design the corresponding test cases for each functional module,and test the performance,function and security of the system according to the test cases we designed.The test results are in our expectation,and the system works normally.