Shot Knowledge Graph Completion

Few-shot knowledge graph completion (FKGC) focuses on predicting missing links in knowledge graphs, particularly for relations with limited training data. Current research emphasizes developing robust models that handle uncertainty inherent in sparse data, often employing graph neural networks, large language models, or normalizing flows to learn effective entity and relation representations. These advancements aim to improve the accuracy and explainability of knowledge graph inference, with applications ranging from automated news generation to more generally improving knowledge base completion in scenarios with limited information. The field is actively exploring methods to incorporate both structural and commonsense knowledge to enhance performance and address the challenges of zero-shot knowledge base completion.