Research On Corporate Service Text Classification System Based On Domestic Policy

In order to build a modern service-oriented government,governments at all levels across the country have added new public government information sections to their official websites and often publish policies in various fields on them,many of which are closely related to business development,such as economic planning,industry standards and winning tenders.Many of them are closely related to the development of enterprises,such as economic planning,industry standards and tender awards.To obtain this information,companies must frequently visit the websites of various government departments,which is time-consuming and labourintensive.Therefore,this paper designs and implements a policy text classification system for enterprises to provide an intuitive and effective policy access and classification service for domestic enterprises and talents.Text classification is an important function of the system and is investigated in this paper.Firstly,this paper analyses the special characteristics of policy text data,and investigates several popular algorithms in the field of deep learning text classification,including a separate algorithm and an algorithm that fuses two algorithms.The separate algorithms include Text CNN(CNN full name Convolutional Neural Networks)and LSTM(Long-Short Term Memory),and the fusion algorithms include Text CRNN(CRNN full name Convolutional Recurrent Neural Networks),TC-LSTM(Text CNN-LSTM)and ATT-CNN(Attention-based Text CNN).After analyzing the characteristics of policy text data and classification algorithms,it is predicted that the model containing Text CNN is more suitable for the policy text classification in this paper.From the experiments,it is clear that Text CRNN far outperforms other algorithms in terms of model performance and learning speed.Therefore,Text CRNN is chosen as the policy text classification model in this system.In order to verify the generality of the Text CRNN model,this paper was conducted on different corpuses and text classification tasks.The corpuses include government policies dataset at all levels dataset and type classification policies dataset in Shandong province provided by a domestic company and provincial government policies dataset in Shandong province and provincial government policies dataset in Guangdong province crawled from the official government website.The classification tasks of the system include policy filtering and policy type classification.The experimental results show that Text CRNN performs well on several corpuses and text classification tasks.After the model training was completed,the web system was designed and implemented in this paper.In terms of technology,the system uses the Django framework based on the python language to deploy the whole system,the front-end uses html5 technology and Bootstrap3 frontend framework to optimise the interface,and the My Sql database for data storage.In terms of functionality,the system implements user registration and login,policy acquisition,policy classification and policy download functions,where the policy classification function includes two sub-functions of policy filtering and policy type classification.Users can independently select the government websites to be crawled,and the system will filter policies that are not related to enterprises,and then automatically classify them into three types: policy,declaration and technology,and finally users can download the crawled policies and the classification results.After rigorous testing,the system basically achieves the expected goals and can meet the needs of users,and the response time of each function is controlled within a reasonable range.