We lack an overall context (e.g., ontology) for MusicBrain database.

[candlecao]

For NLQ2SPARQL, the current experiment is only based on creating context (ontology) for each isolated Natural language question.
See:
https://github.com/DDMAL/linkedmusic-queries/blob/main/LLM2SPARQL_TheSession_WikiData_MusicBrainz/Leveraging%20ChatGpt%20for%20Natural%20Language%20Query%20to%20SPARQL%20on%20Virtuoso.pptx

We need an overall context for further experiment.
Ideally, the context should be referential to converting any natural language question to SPARQL

[candlecao]

I refer to the solution for this DDMAL/linkedmusic-datalake#88

[candlecao]

As of 2024 Dec 10, we have learned that, for NLQ2SPARQL against an RDF database, the Schema rendered by ShEx is more effective than that by OWL (or RDFS). [see: https://github.com//discussions/27]The former serves as a more appropriate context to train LLMs. So, the solution for this issue can be deferred.