Research On The Influence Of COVID-19 Event On The Netizen Emotion Under The Microblog Environment

Since the outbreak of COVID-19,microblog as an important carrier of information transmission,has played an indispensable role in the release and dissemination of epidemic information.The in-depth analysis of the epidemic information and the impact on the netizen emotion contained in COVID-19 event can help governments grasp the public opinions on the Internet as well as do a well work in epidemic prevention and public opinion guidance efficiently.Therefore,this paper studies the three aspects of epidemic event portrait construction,netizen sentiment portrait construction and impact analysis.The specific work is as follows:1.During the construction of epidemic event portraits,we first crawled 26,478 pieces of news related to the new crown pneumonia epidemic between December 1,2019,and April 30,2020 from microblog,and constructed epidemic events through data cleaning and manual annotation.The corpus was extracted and classified according to the four themes of donation,prevention and control,hero and clinical.Then,use the Word2 vec model to train the event extraction corpus to obtain the vector representation containing the semantic information of the epidemic event.Finally,the BILSTM-CRF model is used to extract semantic features to complete the extraction of epidemic events.Among them,the accuracy rate,recall rate,and F value of the four types of themed epidemic events are all above 80%.The time,location,and character identification of the epidemic event elements are relatively good.The accuracy rate,recall rate and F value are all above 70%.The F-value is higher,reaching 80.8%.2.In the process of constructing netizens' emotional portraits,we first crawled 300,000netizens' comments on news related to the new crown pneumonia epidemic between December 1,2019,and April 30,2020 from microblog,through data preprocessing and manual annotation.,and constructed a corpus of netizen sentiment analysis.Then,using the Chinese emotional vocabulary ontology database,the modifier dictionary and the emoji dictionary constructed by the point mutual information method combined with the network text characteristics,the text emotional intensity calculation is calculated,and the netizen emotion are divided into: optimistic,beautiful,sad,angry according to the calculation results.,fear,disgust and fright.Finally,the two-way recurrent neural network model(ATT-BIRNN)with integrated attention mechanism is used to extract the features of netizen emotion,and the highest F value of netizens' emotion recognition reaches 91.3%,and the validity of the ATT-BIRNN model for the analysis of netizen emotion is verified.In the comparative experiments of ATT-BIRNN model,the F value of the model is the highest of 90.9%.3.In the process of impact analysis,the BERT-ATT-BILSTM model is used to learn the semantic information of the correlation between epidemic events and netizen emotion,and the dynamic changes of netizen emotion during the development of epidemic events are obtained,and the experimental results are objectively analyzed.In the comparative experiment to verify the effectiveness of the BERT-ATT-BILSTM model,the accuracy,recall rate and F value of the correlation analysis of the epidemic events on netizen emotion all reached more than 83%,and the F value of the "hero" theme event reached 83%.86%.