Research On Domain Knowledge Graph Technology

As the core of the new knowledge engineering,knowledge graph has become the infrastructure for the application of artificial intelligence in many fields,which has received abroad attention from academia and industry.In recent years,the integration of knowledge graph technology with various industries has become an important trend,and domain knowledge graph as a focus of knowledge graph in specific fields emerged.This paper studies the key technologies of constructing knowledge graph,include knowledge modeling,knowledge representation and knowledge acquisition.For the field of artificial intelligence technology and aviation unsafe incident analysis,We built a literature domain knowledge graph and an aviation safety domain knowledge graph based on unstructured scientific document database and structured network database respectively.The method of constructing domain knowledge graph was improved and innovated,and a domain knowledge graph assisted building system is designed and implemented.In terms of knowledge modeling of domain knowledge graph,this paper proposes an up-down mixed pattern construction method to solve the problem that traditional top-down and bottom-up pattern construction methods are difficult to adapt to the characteristics of strong domain knowledge specialization and fast knowledge updating.This method combines top-down(main)and bottom-up(auxiliary)construction modes,which are divided into four steps,namely top-down pattern layer design,data collection and information extraction,bottom-up pattern layer optimization and knowledge connection.In terms of knowledge representation,for the trend analysis task,we proposed a quad-model to represent time related knowledge,it is the expanding of a triple knowledge representation.This enable domain knowledge graph to answer temporal questions directly.Through the application of domain knowledge graph,we built,it realizes the retrieval and recommendation of related knowledge.