Emotion Analysis Based On Deep Learning And Its Application In Public Opinion Analysis System

With the rapid development of Internet technology,especially self-media industries,a large number of user-interactive comments on news events,people,and various products are generated every day on online social media platforms,such as web pages,forums,and blogs.These comments often contain a variety of emotional tendencies reflecting the personal emotions of the users,such as happiness,anger,sadness,and joy,as well as praise,criticism,and neutrality.By analyzing the emotional tendency of subjective information,we can understand the netizens’ opinions regarding a certain event or product.This can in turn provide a basis for public opinion supervision and product recommendation.Emotion analysis usually refers to text emotion analysis,which is often called opinion mining or tendentiousness analysis.Regardless of the application,the essence of emotion analysis is to acquire and process these subjective texts using computational models,and predict their emotional tendency.Based on an analysis of relevant technologies used in network public opinion analysis and emotional tendency analysis,the present study combined information theory,machine learning,deep learning,and other methods to conduct an in-depth researches on emotion analysis technology.Finally,the results were applied to network public opinion analysis.Specifically,the main contributions of this study include the following aspects.(1)An in-depth analysis was conducted of the key technologies involved in emotion analysis and online public opinion analysis,including information extraction and preprocessing,topic discovery and tracking,and emotional tendency analysis.(2)To address the problem that the current sentiment analysis methods ignore the decline in performance caused by negative words,sentence patterns,and other factors,a sentiment analysis method was constructed by integrating sentence patterns,emotion words,emotion objects,and negative words,and the feasibility of this method was confirmed through experiments.(3)The traditional emotion analysis method is labor-intensive,and its performance depends on manually labeled data.Therefore,based on an in-depth study of deep learning methods,this study proposes a self-attention-based bi-directional long short-term memory(Bi-LSTM)model and applies it to emotion analysis.A comparative study proving the effectiveness of this method was also performed.(4)Addressing the problem that existing public opinion analysis systems can not accurately identify user emotion,this study applies the research results of emotion analysis to the network public opinion analysis and supervision system,thereby proving the application value of the proposed research method in network public opinion analysis.