Allow DESCRIBE to return the (asymmetric) CBD

[VladimirAlexiev]

Problem description

We want to extract all statements about ontology terms that are rdfs:isDefinedBy a certain ontology, so as to provide them to LLM for the purposes of SPARQL generation. We want to use DESCRIBE because it traverses blank nodes, so it takes care of patterns like this:

P rdfs:domain [a Class; owl:unionOf (C1 C2)].
C1 rdfs:subClassOf [a owl:Restriction; owl:onProperty p; owl:allValuesFrom C2]

• DESCRIBE of a property P works great
• but DESCRIBE of a class C is problematic, since in rdf4j, DESCRIBE returns the symmetric CBD (SCBD), i.e. all incoming and outgoing props of the node. So it returns all instance triples "X a C". We can filter them out with some post-processing, but in a large database fetching those instance triples takes a long time (just to discard them afterwards).

Preferred solution

Add some feature to control DESCRIBE. The following databases do it with hint:Query:

• Neptune: https://docs.aws.amazon.com/neptune/latest/userguide/sparql-query-hints-for-describe.html
• BlazeGraph: https://github.com/blazegraph/database/wiki/QueryHints

Another option could be to use FROM with a special named graph.

Are you interested in contributing a solution yourself?

None

Alternatives you've considered

No response

Anything else?

No response