Relation Extraction
Overview
Relation extraction finds links such as causes, locations, interactions, and affiliations between entities or events, turning unstructured material into structured data for knowledge graphs and downstream analysis. Common approaches use supervised encoders and joint entity-relation models, while graph methods, large language model prompting or fine-tuning, and ontology constraints address document context, limited labels, and domain-specific schemas. Evaluation spans sentence-level benchmarks such as SemEval and FewRel, document-level datasets such as DocRED, and specialized biomedical, historical, financial, and multimodal settings. A central challenge is reliable generalization: systems may recognize a relation yet miss its exact spans, implicit evidence, or direction, and reported performance can depend on annotation completeness and evaluation rules.
Latest papers
82 papers16 papers in the last four weeks, up 167% on the four weeks before. 0.1% of all new papers.