Interesting and good to know - the way I've handled similar kinds of things before (for SQL), is if you have a list of predicates or entity types you need to pick from you can populate an enum in a JSON schema. If the list is way too long (it probably is), you can get K nearest neighbor of the entity types to the natural language query and populate the enum with K values (where K might be something reasonable like 10 to 20). You'd need to have a dataset of all entity types stored locally to do this of course.

I know this sounds hand-wavy, but I have had good results doing similar things when trying to pick the correct foreign key out of a table with 1000s of rows.

Not for SPARQL, no. But I have done it for other query DSLs like SQL and Opensearch.