Research And Implementation Of Micro Blog Emotion Monitoring System Based On Public Opinion Analysis

With the development of the Internet into a new era and the rapid development of the media industry,Microblog has become a medium for Chinese people to obtain information and express opinions.Microblog contains a lot of text information,such as current events,social phenomena,economic information and comments on it.The use of computer technology to conduct emotional analysis of the text data displayed on Microblog will help relevant departments,enterprises and institutions to accurately understand the public's attitudes and opinions and analyze the impact of events.It can carry out reasonable public opinion analysis and monitoring for relevant departments,enterprises and institutions,and provide accurate reference information for formulating strategies to deal with and solve public opinion.Moreover,the analysis of Microblog public opinions in various fields can help us understand the situation of various industries and their products,which plays a positive role in promoting the development of the industry.This paper studies and implements a micro-blog sentiment monitoring system based on public opinion analysis,removes false comments from this data of micro-blog,and obtains real user comment data,which improves the accuracy of public opinion analysis on this data of micro-blog.PU LEARNING algorithm and LSTM algorithm are mainly used.Combining PU LEARNING model and LSTM model,a kind of emotion analysis based on false comment recognition and deep learning is proposed,which is referred to as PU-LSTM model for short.1.Firstly,PU LEARNING algorithm is used to preprocess and vectorize the acquired micro-blog comment data.2.Then three evaluation indicators and eight judgment attributes are proposed.Status Index,Content Index and Behavior Index,Eight Attributes under PU-LSTM: User Credibility,Timeliness of User Evaluation,Text Length of User Comment,Relevance of Evaluation Content and Topic,Proportion of Sensitive Words in Comment Content,Similarity of Online Comment,Time Interval of user Registration and Intensity of Emotion Expression.3.Based on the above indicators,the PU LEARNING model is used to mark the false comments and eliminate them.4.Finally,based on the text data after eliminating false comments,LSTM algorithm is used to realize the public opinion analysis of the text.In the process of system implementation,API and other functions are used to obtain hot micro-blog and make heat map,and LDA algorithm is used to put forward the keywords of each emotion,which are presented in the form of word cloud.PU-LSTM model can help users to analyze the real user evaluation to obtain a more accurate analysis of public opinion.The experimental results verify that the emotion analysis system designed and implemented in this paper has effectively completed the expected functions.However,this paper has certain limitations.Due to the deviation of the data set,the deviation of the experimental results and the lack of consideration in the algorithm cannot make a more accurate analysis of public opinions.Meanwhile,it is impossible to analyze a series of influences caused by false comments and propose expert system countermeasures for the analysis results of public opinions.