Topic Mining Of Policy Review Text Based On Sentiment Classification

Public opinion plays an important role in the formulation and improvement of public policy.With the development of the Internet,the Internet has become an important carrier of public opinion,and the online expression of public opinion is online comment.Therefore,in order to understand the public's real attitudes and opinions on relevant policies,it is inevitable to mine and analyze online comments on relevant policies.From the perspective of public policy evaluation,this thesis conducts sentiment analysis and topic mining based on online comments of policy,aiming to provide methods and process reference for text mining of online comments of public policy,and give full play to the role of online comments in the process of policy formulation and improvement.This research takes online comments of public policy as the research object.Firstly,it summarizes the application and development status of text mining technologies such as text sentiment classification and text topic mining.On this basis,an analysis model for online comments on public policy network is constructed by using the above methods,which mainly includes three parts: first,data collection and preprocessing.Determine the data source and data collection method,use network information collection technology to collect and store online comments,and use text preprocessing technology to clean and preprocess the collected original comment data.Second,text emotion classification.Based on the sentiment dictionary,this thesis classifies the comments and analyzes the distribution of the commentators' emotional attitudes towards relevant public policies.Thirdly,LDA model is constructed for topic mining of review texts with different emotional tendencies,and topic distribution under different emotional tendencies is analyzed.Using the above analysis model,the "double reduction" policy of education was selected as the case study,and a total of 12,877 relevant comments were collected with weibo comments as the data source.After data cleaning and pretreatment,the remaining 12,366 valid comments were obtained.Combined with How Net dictionary and NTUSD dictionary as the basic emotion dictionary,and based on word2 VEc word vector technology using word similarity calculation to expand the emotion dictionary,based on the emotion dictionary sentiment classification of the comment text to get the emotion distribution.The results show that 24% of the valid comments are positive,21% negative and 55% neutral,and the vast majority of the comments focus on five categories of subjects: children,parents,schools,teachers and policies.LDA theme model is used to mine positive,negative and neutral sentiment comment texts respectively.The results show that the positive emotional comments mainly express support for the "double reduction" policy and affirm the positive effects of the "double reduction" policy.Negative emotional comments mainly express concerns about the "double reduction" policy and the potential adverse effects it may bring.Neutral sentiment comments mainly focus on the details of the implementation of the "double reduction" policy and the conflict with the current educational environment.Finally,this thesis summarizes the above empirical parts,and draws the analysis conclusion: At present,the public's concerns and doubts about the education "double reduction" policy mainly focus on the negative effect of the "double reduction" policy,the implementation of the "double reduction" policy,the details of the implementation of the "double reduction" policy and the prospect of the education and training industry under the "double reduction" situation.Therefore,relevant departments need to make targeted follow-up plans for the above problems,and timely answer questions to the public,so as to dispel the public's concerns and doubts about the policy,improve the public's confidence in the policy prospect,and lay a good public opinion foundation for the implementation of the policy.