Design And Implementation Of COVID-19 Network Public Opinion Monitoring System Based On Deep Learning

With the development and maturity of information technology,the Internet has made great progress in the direction of precision and intelligence,and has the characteristics of information diversity,instantaneity and universal interaction.Since the sudden outbreak of COVID-19 in December 2019 and its subsequent spread across the world,the topic remains a hot topic of global attention today.Online social media platforms,represented by Sina Weibo,spread public opinion on COVID-19 with netizens' emotions rapidly.Negative emotions can easily cause panic and affect the Internet ecology and social stability.Therefore,timely grasp the trend of public opinion,correctly guide public opinion emotion become the top priority of stability maintenance work.With the popularization and development of deep learning technology,computers can process text data more efficiently and gradually replace the manual mode of public opinion monitoring.In this paper,an online public opinion monitoring system based on deep learning has been implemented.The main work of this paper includes the following aspects:(1)Data collection and preprocessing: In view of the lack of COVID-19 emotion data sets,this paper designed and implemented the crawling of COVID-19 public opinion related microblogging text by adopting Scrapy distributed crawler framework.The data set was annotated with sentiment orientation,and data pretreatment methods such as regular expression,jieba word segmentation and Hit stop word table were used to clean,divide and remove the original data,and finally the data needed for sentiment orientation recognition were obtained.(2)Identification of emotional orientation: In view of the characteristics of different lengths and multiple popular terms in microblog text,this paper proposes an identification model of emotional orientation based on Bert-Bilst M in order to better complete the monitoring task of public opinion.The BERT(Bidirectional Encoder Representations from Transformers)pretrained language model can generate richer semantic information from multiple dimensions.The Bidirectional Long Short Term Memory(BiLSTM)model performs better in feature extraction of text context,and finally outputs the emotion orientation recognition results after the full-connection layer and Softmax classifier.The Bert-Bilst M model combines the advantages of the two models and can extract local and global feature information at the same time.Compared with other deep learning models,this model is proved to be effective and superior in the task of emotion orientation recognition.(3)System implementation: Through the demand analysis of public opinion monitoring system,select system development tools,design and implement the system.The online public opinion monitoring system mainly includes user registration and login,public opinion overview,emotional tendency distribution,public opinion word cloud display,etc.Each function module is presented to the user in the form of visual charts to improve the overall user experience.