Multilingual Relation Extraction: A Survey

Abstract

Relation extraction plays a fundamental role in applications of various research fields such as knowledge graph construction, event extraction, and question answering over knowledge graphs, as they often rely on extracting relationships between named entities. Relation extraction has been extensively studied in high-resource languages like English. However, there remains a significant gap in supporting languages with limited resources, defined as those lacking comprehensive annotated corpora, linguistic tools, or pre-trained models, limiting the completeness and accuracy of applications that rely on multilingual data. This paper provides a comprehensive survey of recent advances in relation extraction, focusing on multilingual approaches. We systematically review state-of-the-art methods, datasets used for evaluation, and key features leveraged in these approaches. Additionally, we perform a detailed comparative analysis of the surveyed methods, examining their methodologies, target domains, levels of extraction, explored languages, and effectiveness. Finally, we identify promising directions for future research, with an emphasis on enhancing multilingual relation extraction.