Research On The Construction And Application Of Knowledge Graph In The Military Field

The web data is growing by leaps and bounds,however,the vast amount of knowledge hidden in the data has not been fully explored.In order to be able to make efficient use of the data,research scholars and numerous companies at home and abroad have begun to organize this information by constructing a knowledge graph.Knowledge graph combines concepts,entities,events,and their relationships that exist in the real world together and forms a semantic web diagram,and expresses the data information in a form that more closely resembles the human cognitive world.At present,existing knowledge graphs are mainly oriented towards generic fields and there is not a good methodology to construct and represent knowledge graph for specific purpose field,especially in the military domain,where the research process for knowledge graph is slow.As a result,this dissertation aims to study how to construct a comprehensive and stable knowledge graph for the military field and how to use it efficiently for applications such as knowledge retrieval and question and answer.This dissertation investigates how to construct knowledge graph in the military field and the application implementation of military knowledge graph.Open military data will be used efficiently by applying the techniques of knowledge graph to the military field.First,this dissertation uses a crawler framework to collect structured and unstructured data from sites such as HDWiki and Global Military,and then process and supplement the data,and then conducts domain ontology model based on the classification of these data and the intrinsic relationships between the elements.Secondly,this dissertation completes the military knowledge extraction with two subtasks of entity recognition and relationship extraction.In this dissertation,we firstly used the KNN algorithm to classify military entities,constructed a military entity library containing more than 79,000 entities,and the BiLSTM-CRF algorithm was used to complete the military entity recognition task and the PCNN algorithm was used to extract relationships between the entities to obtain the triangulation information.Using the extracted knowledge,a military knowledge graph with a clear hierarchy and flexible structure is constructed.Finally,this dissertation adopts Neo4 j graph database to store data,and develops and implements six functions of identification and query of military entities,relationship query,military knowledge overview,military knowledge Q&A and image retrieval on the constructed military knowledge graph,which provides smarter and more comprehensive feedback on the military information that users interest.