Design And Implementation Of Intelligent Text Circulation System For Government And Enterprise Based On Deep Learning

The impetuous development of communication and Internet technology has spawned the demand for all kinds of government affairs online and made the number of government affairs data increase explosively.At the same time,the promotion of "big data" has spawned the demand of government departments for data governance and analysis.Strengthening data empowerment has become one of the key tasks of government departments.A government big data analysis platform with "gathering,sharing and using" as the core is essential.At present,due to the characteristics of huge,complex,high noise,numerous categories and unequal data distribution of government data,the development of the topic of government and enterprise text data analysis and processing is relatively slow,At the same time,there are few researches on e-government text classification and clustering.Government and enterprise data are usually stored in the form of text.In the process of government affairs text processing,how to maximize the use of data resources in government affairs text by text classification and text clustering and other methods,to ensure the efficiency of government affairs work,is an urgent problem to be solved.In recent years,the impetuous development of deep learning technology has provided strong support for Natural Language Processing(NLP),and many relevant researchers have begun to use deep learning for text analysis.Deep learning technology has been applied to word embedding representation,machine translation,text classification,text clustering and other scenarios,and has achieved convincing results.Combined with the performance of deep learning in text analysis tasks and the analysis needs of government-enterprise texts,this thesis designs an intelligent flow system for government-enterprise texts.This thesis first introduces the status quo and necessity of government and enterprise text data analysis,analyzes the role that deep learning technology can play in data analysis,and explains the practical application of government and enterprise text data analysis.Various related technologies involved in the operation and implementation of the government-enterprise text intelligent flow system are also introduced in detail.From various processing methods of text representation to the technical introduction of text classification and text clustering model,the addition of attention mechanism can play a role in strengthening the characteristics of text classification model.In addition,other development technologies related to system data storage are introduced,such as My SQL database development technology,Elasticsearch distributed search engine,API interface development technology and data enhancement.Secondly,an algorithm based on Bi LCNN and word embedding is proposed,which combines attention mechanism to classify text.Firstly,BERT model is used to express the text information of government affairs,and the features are extracted by combining Bi LSTM and CNN.Simultaneously introducing attention mechanisms,combining temporal and local features to enhance features.Finally,Softmax is used to sort the texts.Experiments show that the accuracy of BERT word embedding in the hybrid model is improved by 3.9% and 2.51%compared with CNN and Bi LSTM models,respectively.Then the event public opinion analysis clustering model based on DBSCAN and similarity calculation is introduced.The similarity relationship between texts is determined based on distributed Elasticsearch engine and text similarity calculation principle,which determines whether data can be clustered into the same class.On the basis of the completion of text content clustering,DBSCAN clustering algorithm is used to calculate the distance between geographical locations of similar data to determine whether it is the same event.This is of great significance to the results of public opinion analysis of the event and has achieved good results.Finally,the Integral design and realization process of the intelligent flow system of government and enterprise text is introduced,and the sample of data set in the operation of the system is introduced.In practice,from the aspects of improving operation efficiency,improving accuracy and so on,the operation process of the system is designed reasonably,and the result is evaluated reasonably.