Knowledge graph completion method based on global and local comparative learning

• Main Authors: MA RUIXIN, WANG XIAORU, WANG LONGFEI, CHEN ZIXUAN, BU XIYA
• Format: Patent
• Language: Chinese; English
• Published: 08.03.2024
• Subjects: CALCULATING; COMPUTER SYSTEMS BASED ON SPECIFIC COMPUTATIONAL MODELS; COMPUTING; COUNTING; ELECTRIC DIGITAL DATA PROCESSING; PHYSICS

The invention belongs to the technical field of knowledge graph completion, and provides a global and local comparative learning-based knowledge graph completion method, which comprises the following steps of: firstly, constructing a local multi-mode context encoder, and developing three attention modes by utilizing a topological structure of a knowledge graph to model context information of an entity so as to obtain local semantic features of the entity; a global attribute consciousness encoder is constructed in parallel, an entity-entity attribute interaction graph is constructed based on the whole knowledge graph to obtain the interaction strength between two entities, and a lightweight graph aggregation network is designed to be combined with the attribute interaction graph to update the semantic features of the entities, so that the global semantic features of the entities are obtained. And then carrying out comparative learning between the local features and the global features of the entities, adopting