Design And Implementation Of Administrative Cases Quick Handling System Based On Natural Language Processing

In recent years,in order to realize the systematic and standardized management of law enforcement behavior,general offices have distributed?Opinions on deepening the standardized construction of public security law enforcement behavior?and?Regulations on the procedures for public security organs to handle administrative cases?,and put forward the regulations for the implementation of the quick handling procedures for cases.The procedure is mainly aimed at the phenomenon of more cases and less police,and its application can effectively reduce the work pressure of police,also improve the efficiency of case handling and reduce the cost of law enforcement.Combined with the application needs of quick handling procedure and the construction needs of intelligent police systems,this paper design and implement a cases quick handling system based on natural language processing,which assists the police in three processes:evidence collection,case handling and administrative punishment.In the development of the whole system,the main work of this paper is realizing three subsystems:record making,semantic analysis and administrative penalty.Among them,the record making subsystem relies on the dialogue management algorithm to inquire the record,correct wrong words and detect the consistency of questions and answers,making record quickly.The semantic analysis subsystem extracts key elements in the record,so as to facilitate the police to read.In the intelligent punishment subsystem,there is a complete knowledge graph of punishment,in order to making corresponding punishment decisions for the violators according to the actual situation.In this project,the front-end is developed with Vue.js.The back-end is developed with Java and the system architecture is Spring MVC.The algorithms are developed with Python,and Tensorflow is used as the deep learning framework.After training,the weight files are stored and encapsulated in Docker to provide external services.The record making subsystem is implemented with multi-round dialogue management dominated by system,and the text correction algorithm is based on KenLM language model,the text classification algorithm is based on BERT model,the algorithms are used to assist in the record compliance detection.In the semantic analysis subsystem,the named entity recognition algorithm is based on BiLSTM+CRF,used to identify the key elements in the record.In the intelligent punishment subsystem,Neo4j database is used as the repository of punishment knowledge graph,and the query conditions are transformed into Cypher language for query operation.At present,the three subsystems have been developed and put into use in police station,and have handled a number of administrative cases efficiently and quickly,achieved a good result.