Research On Public Opinion Evolution Of Winter Olympics Microblog Based On Emotion Classification And Topic Mining

Since China hosted the Beijing Olympic and Paralympic Winter Games in February2022,"Winter Olympics" has become one of the hot topics in microblogs,and the attention of various sectors of society to the distribution and reuse of Olympic heritage,the development of snow sports and other related topics has increased significantly.On microblogs social media platforms,the content of microblog and users’ behaviors,such as forwarding,commenting,and liking,infect users’ emotional attitudes and topic trends usually,which directly promote the fermentation,evolution,and dissemination of public opinion,and have an impact on the governance of online public opinion and social security.In this paper,the data source is the Winter Olympics-related microblog text.We focus on extracting the hidden information of microblog text through emotion classification and topic mining methods,explore how to improve the text classification performance of the microblog public sentiment under resource constraints,and how to integrate the microblog users’ behaviors into the public opinion topic mining.After visual display and interpretation of the evolution of the sentiment and topic of microblog public opinion in the Winter Olympics,we summarize the evolution law of public opinion,which will form a certain reference for the public opinion management of different types of events on different platforms.The main work of this paper is divided into the following three aspects:Firstly,an improved emotion classification model based on Knowledge Distillation method is proposed.Based on the Knowledge Distillation method,a weighted triple distillation loss function is proposed to improve the emotional prediction performance of simple neural networks at its low model complexity level.The experiment results show that our method is helpful for deploying emotional analysis under resource constraints and applicable to the deployment of other natural language processing tasks.Secondly,a microblog topic mining method integrating users’ behaviors is proposed.This method quantifies the communication influence of forwarding,commenting,and praising behaviors used communication contribution,and integrated it into the text topic mining process.Then,an online questionnaire is designed based on the user content evaluation perspective,and the reliability of topic recognition results is evaluated by the Analytic Hierarchy Process method.The results show that the topic recognition results integrated with users’ behaviors are more reliable.Our method crossed the research of social network analysis and text mining method,and provide a more comprehensive analysis idea for public opinion analysis.Finally,based on the sentiment classification and the topic mining methods we proposed,combining the visual displays of the number of microblog posts,the behavior of microblog users,public opinion emotion,and microblog topic around the Beijing Winter Olympics,we interpreted the distribution and reasons for the changes in public opinion hotspots on microblogs around the Beijing Winter Olympics and summarized the evolution low of public opinion on the microblog for the Winter Olympics,which will conducive to providing decision support for the development of related industries in the post-epidemic era and the post-Winter Olympics era.