This dataset accompanies the paper AI-Assisted RDF Authoring from JSON Data.
It contains the metal-organic framework (MOF) synthesis JSON input used in the paper, the natural-language hints submitted to MetaConfigurator's AI assistant, the RML mappings generated from those hints, the RDF (JSON-LD) results obtained by executing the mappings, the SHACL shape used to validate the successful result, and the prompt, generated query, and CSV result of the AI-assisted SPARQL workflow.
The artifacts document three RML-generation runs and one SPARQL run: a minimal hint with gpt-4o (valid but shallow mapping), the same detailed hint with gpt-4o (retained as a failure case: unconditional parent-map links cause Cartesian products and drop required relations) and with gpt-5-mini (the successful case used in the paper), and a natural-language SPARQL request answered by gpt-5-mini. The prompts contain natural-language modeling instructions but no worked RML/Turtle snippets.
The deposited files are the exact artifacts evaluated in the paper. Model outputs can vary across invocations and service versions. See README.md for the full file inventory and reproduction instructions (including pySHACL validation).