Design And Implementation Of Public Opinion Monitoring System Based On Deep Learning Technology

With the development of the Internet,the use of various media and social platforms has become an important way for people to track hot spots and obtain information.However,the freedom and suddenness of network platforms have caused some negative or untrue information to spread rapidly in the network,which has had a negative impact on the government and enterprises.Therefore,how to regulate the network platform reasonably and effectively is an urgent problem to be solved.For enterprises,the supervision of public opinion information can not only find out the negative information related to the brand in time,maintain the image and reputation of the enterprise,but also optimize the product and management by mining user feedback,so as to improve the core competitiveness of the enterprise.At present,the relevant public opinion monitoring system was still in the stage of development,in terms of system integrity,applied some deficiencies,at the same time,with the development and application of deep learning technology,unstructured text data can through the computer to recognize and deal with more and more,in a more efficient way to replace the original manual supervision.By sorting and analyzing the text data in the network platform,this paper analyzes and designs the public opinion monitoring system,which can meet the business needs of enterprises for brand monitoring and user opinion mining.The system requirement analysis stage of the research and reference to the current mainstream of the monitoring system,the system function module division,mainly including brand monitoring module,hot topic module,warning module and system management module,enables the system to convenient for the brand information monitoring,can also be hot topic by looking at the attention it is happening or potential problems,also can use the system through the custom early warning scheme received warnings that,look at content and so on.In the implementation stage of the system,through data analysis and algorithm research,this paper selects the BTM model applicable to the short text for text modeling,generates a topic model,and selects the number of topics by topic similarity,and completes the selection of hot topics by topic importance calculation.Meanwhile,the mixed model of emotion classification combined with Text CNN and GRU was constructed to complete the emotion classification of the short text.The algorithm is implemented and a complete public opinion monitoring system is realized.The realization process of the system is described in detail through the flow chart,class diagram and sequence diagram.Finally,functional and non-functional tests were carried out on the public opinion monitoring system to verify that the system could meet enterprises' demands for monitoring their brands and improve enterprises' ability to control public opinion risks.