Construction of hierarchical knowledge graph based on deep learning

Abstract

With the continuous deepening of knowledge graph research, more and more knowledge is softened together, and the knowledge in professional fields is also emerging. Although people can quickly identify the knowledge they need based on their needs, machines cannot. There are many problems in the organization and application of traditional graphs, such as the inaccuracy of knowledge representation, which makes it difficult to obtain. The lack of clear knowledge layers causes a lot of irrelevant knowledge to appear after the query. The chaotic structure of knowledge in the graph causes query time-consuming. Therefore, considering the different layers of knowledge representation and the knowledge used to solve complex engineering problems, we propose to divide knowledge into three layers - basic knowledge, deep knowledge, and application knowledge and an agent-based hierarchical knowledge graph construction framework and methodology. The deep learning model method is used in the classification agent to realize the automatic division of knowledge type, pass the classification results to the corresponding knowledge agent. This knowledge agent is able to construct the hierarchical knowledge graph by the same layer knowledge or cross-layer knowledge. This method of constructing the hierarchical knowledge graph has practical significance in the application of the knowledge graph, which makes the knowledge graph have a wider application and practical value.