Enabling Knowledge Graph Understanding at Scale with the EXplore Your Graphs ENgine (EXYGEN)

RESUMO

EXYGEN enables conversational access to knowledge graphs at scale by combining ShEx schemas, retrieved triples, and example question-query pairs in a retrieval-augmented text-to-SPARQL pipeline, reaching an exact match of 0.419 on SciQA without LLM fine-tuning. A predicate-coverage-aware parallel graph sampling strategy generates the structured metadata this relies on from very large KGs, retaining high predicate coverage with minimal triple loss and reducing runtime by over 80x.