arXiv 2511.05991
Ontology Learning and Knowledge Graph Construction: A Comparison of Approaches and Their Impact on RAG Performance
By Tiago da Cruz, Bernardo Tavares, et al.
Published 2025-11-08
Citation lineage
Review the prior work and downstream research connected to this paper.
Retrieval-Augmented Generation (RAG) systems combine Large Language Models (LLMs) with external knowledge, and their performance depends heavily on how that knowledge is represented. This study investigates how different Knowledge Graph (KG) construction strategies influence RAG performance. We compare a variety of approaches: standard vector-based RAG, GraphRAG, and retrieval over KGs built from ontologies derived…