Build2SPARQL Benchmark for Building Knowledge Graph Text-to-SPARQL Querying
October 2, 2026
Build2SPARQL introduces a large-scale benchmark for translating natural language into SPARQL queries within building automation systems. The dataset uses a KG-grounded pipeline that validates queries via graph-traversal code and leverages LLMs solely for generating natural-language questions across six query-pattern families.
HOW THIS AFFECTS YOU
●
builderThis provides a standardized way to test language agents meant for building management systems.
●
researcherYou can use this to evaluate text-to-SPARQL accuracy in domain-specific semantic knowledge graphs.