METHOD AND APPARATUS FOR LEARNING KNOWLEDGE REPRESENTATION, DEVICE, STORAGE MEDIUM AND PROGRAM

摘要

An embodiment of the present application discloses a knowledge expression learning method, an apparatus, an electronic device, a storage medium, and a program, and relates to the field of natural language processing, deep learning, and knowledge graph technology. One specific implementation method of the method includes the steps of sampling a knowledge graph subgraph from a knowledge base; Obtaining a sequencing text by sequencing the knowledge graph subgraph; And reading the ranked text according to the order in the knowledge graph subgraph using the pretrained language model, and learning and obtaining the knowledge representation of each letter in the ranked text. Knowledge expression learning in this implementation method is about learning entity and relation expression in the knowledge base. In a low-dimensional space, the semantic association between entities and relations is efficiently calculated, data sparse problem is effectively solved, and knowledge is acquired. And can significantly improve the performance of fusion and reasoning. Using the powerful knowledge acquisition and context analysis capabilities of the pretrained language model, the knowledge expression learned in the pretrained language model can better represent the complex relationships in the knowledge base.