{"slug": "knowledge-graph-based-retrieval-augmented-generation-for-schema-matching", "title": "Knowledge Graph-based Retrieval-Augmented Generation for Schema Matching", "summary": "A new arXiv paper (2501.08686v2) introduces KG-RAG4SM, a knowledge-graph-based retrieval-augmented generation method for schema matching that improves precision by 39.79% and F1 score by 30.56% over LLM-based state-of-the-art methods such as Jellyfish-8B on the MIMIC dataset. On the Synthea dataset, KG-RAG4SM reports gains of 69.20% in precision and 21.99% in F1 over pre-trained language model baselines such as SMAT, without requiring re-training. The method combines LLM-based, vector-based, graph traversal-based and hybrid retrieval with reranking to pull relevant subgraphs from external knowledge graphs, and its instruction-tuning knowledge exclusion component is designed to keep end-to-end schema matching truthful despite suboptimal KG context.", "body_md": "arXiv:2501.08686v2 Announce Type: replace-cross \nAbstract: Traditional similarity-based schema matching methods fail to resolve semantic ambiguities and conflicts in complex, domain-specific mappings due to a lack of commonsense and domain-specific knowledge. In parallel, large language models (LLMs) without sufficient background knowledge tend to hallucinate, further complicating schema matching. To overcome these limitations, we introduce KG-RAG4SM, a Knowledge Graph-based Retrieval-Augmented Generation for Schema Matching. This method integrates novel retrieval mechanisms, including LLM-based, vector-based, graph traversal-based, and hybrid techniques with advanced reranking schemes to extract the most relevant subgraphs from large external KGs. Our approach achieves superior results without requiring re-training, as evidenced by improvements of 39.79% in precision and 30.56% in F1 score over LLM-based state-of-the-art (SOTA) methods (e.g., Jellyfish-8B) on the MIMIC dataset, and enhancements of 69.20% in precision and 21.99% in F1 score over pre-trained language model (PLM)-based SOTA methods (e.g., SMAT) on the Synthea dataset. Additionally, KG-RAG4SM scales effectively across KGs from commonsense KGs to domain-specific ones, while its instruction-tuning knowledge exclusion component ensures truthful performance in end-to-end schema matching tasks despite the suboptimal relevance of KG context. Our empirical analysis and case studies from real-world scenarios show that the hallucination problem in LLMs for schema matching is well mitigated by our solution.", "url": "https://wpnews.pro/news/knowledge-graph-based-retrieval-augmented-generation-for-schema-matching", "canonical_source": "https://www.machinebrief.com/news/knowledge-graph-based-retrieval-augmented-generation-for-sch-r1si", "published_at": "2026-11-09 05:00:00+00:00", "updated_at": "2026-11-09 06:21:25.460105+00:00", "lang": "en", "topics": ["artificial-intelligence", "large-language-models", "ai-research", "natural-language-processing", "ai-safety"], "entities": ["KG-RAG4SM", "Jellyfish-8B", "SMAT", "MIMIC", "Synthea", "arXiv"], "also_reported_by": [], "alternates": {"html": "https://wpnews.pro/news/knowledge-graph-based-retrieval-augmented-generation-for-schema-matching", "markdown": "https://wpnews.pro/news/knowledge-graph-based-retrieval-augmented-generation-for-schema-matching.md", "text": "https://wpnews.pro/news/knowledge-graph-based-retrieval-augmented-generation-for-schema-matching.txt", "jsonld": "https://wpnews.pro/news/knowledge-graph-based-retrieval-augmented-generation-for-schema-matching.jsonld"}}