What is KG completion?

KG completion (Knowledge Graph completion) is the task of filling in missing knowledge in a knowledge graph. A knowledge graph stores entities and relationships between them, but often the data is incomplete or contains gaps. KG completion uses machine learning methods and logical inference to predict missing edges (relationships) or nodes (entities) to make the graph more complete and useful.

For example, if the graph contains information that "Alice works at Company X," but there is no connection between "Alice" and "city," KG completion can predict that Alice lives in city Y based on similar data or patterns.

Main approaches:

• Embedding models for entities and relationships (TransE, ComplEx, etc.)
• Graph neural networks (GNN)
• Logical inference and rules

KG completion helps improve the quality of recommendations, search, and data analysis in systems that use knowledge graphs.