Research On Intelligent Work Order Dispatch With Enhanced Public Opinion Topic Recognition Function In Political Situation And Public Opinion

The origin of political situation and public opinion can be traced back to the ancient ’ peopleoriented thought ’.The attention and application of public opinion by ancient Chinese rulers originated from their ’ people-oriented ’ thought.E-government service hotline is an important channel to reflect the political situation and public opinion,reflect the problems of citizens and cities,and solve the needs of citizens government services.In this process,the rational use of government service hotlines in e-government resources is emphasized to improve the service response to citizens and the efficiency of problem solving.Under the new historical conditions,’ Internet + government service ’has gradually become the government ’s public service means with ’ Internet + ’ as the main body.Through data analysis and mining,intelligent means is used to show the people ’s livelihood demands and regional distribution expressed by the public through the government service hotline,to understand the problems that the public urgently need to solve and the main sources of urban contradictions,and to optimize the process of the government service hotline to form a good social governance process and pattern.Based on the extensive review and analysis of domestic and foreign research literature and data on political situation and public opinion,LDA topic model is proposed by the dissertation to extract the main public opinion topics in political situation and public opinion,designs the empirical analysis of confusion degree and topic consistency index,extracts the number and content of topics in political situation and public opinion,studies the time distribution law and spatial distribution pattern of different topics,and verifies the effectiveness of topic model and index in political situation and public opinion.Aiming at the problem that the original appeal work order classification effect is not good,it is proposed to expand the work order delivery research into a multi-dimensional feature extraction problem,and design the LDA unsupervised learning model and BERT pre-training feature fusion method to realize the potential spatial representation of the appeal information.The LDA model is used to obtain the topic information in the political situation and public opinion and the BERT pre-training model is used to encode the semantic representation information in the political situation and public opinion.Based on the feature fusion of LDA＿BERT,the text features are input into softmax,KNN and SVM classifiers for model training and work order dispatch.The experimental results show that compared with the original baseline classification model,the classification model of LDA＿BERT feature extraction and support vector machine proposed in this paper has improved the accuracy,precision,recall rate and F1 value in the intelligent work order delivery task in the field of political situation and public opinion,which verifies the effective improvement of LDA＿BERT feature extraction and SVM in the work order delivery model.