Design And Implementation Of An Entity Relationship Extraction Module For A Public Security Inquiry Record Management System

With the continuous improvement of the national legal system in the new era,relying on the ever-changing Internet technology,the public security system gradually adopts the network platform to manage the recorded information.However,to clear up criminal cases,further excavation of the recorded information is required.At present,the case sorting relying on human’s efforts is inefficient and the integration of all effective information is significantly difficult.The entity relationship extraction can extract the attributes of the persons involved in the case and the relationship among them and then construct the relationship network of these persons,thereby assisting the public security system in cracking the case.It can be seen that the development of entity relationship extraction module is of great significance.After investigating the research status of entity relationship extraction and combining the transcript information given by the public security system,this paper chooses two methods,which are based on template and neural network respectively,to implement entity relationship extraction.After the extraction is completed,a graph database is used to visualize the extraction results.First,this article introduces the research background of entity relationship extraction,and analyzes the main problems existing in the task of relationship extraction.Then,the author performs the requirement analysis and outline design of the entity relationship extraction part,determines the functional requirements and non-functional requirements of this part,and then designs the overall architecture.Next,this article describes the detailed design and implementation of the entity relationship extraction module.The realization of entity relationship extraction module is further divided into template-based entity relationship extraction and neural network-based entity relationship extraction.The template-based entity relationship extraction is divided into three steps.The first step is to complete the name correction of the word segmentation,the second step is to resolve the corrected word segmentation result,and the third step is to fill in the character attribute slot and extract the character relationship according to the resolution result.The neural network-based entity relationship extraction method,after comparing PCNN(Piecewise Convolutional Neural Networks)model,GRU(Gated Recurrent Unit)model and BiLSTM(Bidirectional Long Short-Term Memory)model performance in entity relationship extraction tasks,optimizes the best performing PCNN neural network.Finally,illustrate the result of the module test.This paper studies the transcript information in detail,and realizes the entity relationship extraction model to construct a network of persons involved in the case,which is convenient for the case-handling personnel to analyze and sort the case.At the same time,this article also promoted the development of entity relationship extraction technology in the field of public security to a certain extent.