directSubClassOf, directType, directSubPropertyOf

[VladimirAlexiev]

(This is an addition to the RDFS vocab not a SPARQL feature, but I still think this group is an appropriate forum to post such addition.)

Why?

rdfs:subClassOf, rdfs:subPropertyOf are reflexive and transitive, and rdf:type is transitive over subClassOf.

But reflexivity is most often useless, and having the direct (step, non-transitive) variants of such properties is useful in many situations, eg:

• to display a tree of classes
• for other detailed subclass work (eg see https://github.com/edmcouncil/fibo/issues/1187)
• to implement transitive reasoning more efficiently (eg see https://graphdb.ontotext.com/documentation/enterprise/rules-optimisations.html#transitiveover)

Previous work

• rdf4j has props sesame:directSubClassOf, directType and directSubPropertyOf:
  • https://rdf4j.org/javadoc/3.1.0/org/eclipse/rdf4j/model/vocabulary/SESAME.html
  • http://archive.rdf4j.org/javadoc/sesame-4.1.0/org/openrdf/sail/inferencer/fc/DirectTypeHierarchyInferencer.html
  • https://graphdb.ontotext.com/documentation/standard/query-behaviour.html#how-to-use-rdf4j-direct-hierarchy-vocabulary
  • Note: https://github.com/eclipse/rdf4j/issues/2460 is a task by @jeenbroekstra to move them to an rdf4j namespace.
• Jena has ReasonerVocabulary, which includes directSubClassOf "The property that represents the direct/minimal version of the subClassOf relationship", directSubPropertyOf and directRDFType.
  • These are Java parameters not RDF props
  • @afs are there RDF props to access these? Or it depends on reasoner? What RDF props does the Jena reasoner make available?

Workarounds

• One could hack it with complex and inefficient queries like this:

  select * {
  ?x rdfs:subClassOf ?y
  filter (?x != ?y && 
  !exists {?x rdfs:subClassOf ?z. ?z rdfs:subClassOf ?y. 
    filter(?x != ?z && ?z != ?y)})}

• an approximation is also possible by using Entailment Regimes or vendor specific mechanisms like onto:explicit, eg

  
  # Q1
  select * from onto:explicit {
  ?x rdfs:subClassOf ?y
  }
  # Q2
  select * from onto:explicit {
  ?x rdf:type ?y
  }

But this doesn't work in some situations:
- if a transitive relation is inserted explicitly (I'm not sure whether directSubClassOf eliminates such relation instance, either)
- in Q2, if rdf:type itself is inferred (eg from rdfs:domain)

## Proposed solution

It would increase interoperability if this group can pick a w3c namespace for such properties and standardize the names, and rdf4j and jena agree to expose their props under these standard names

## Considerations for backward compatibility

It would be best if rdf4j and jena continue to support these props under both their current names, and the unified names. I hope that such "aliasing" of magic props is easy to achieve.

[afs]

@afs are there RDF props to access these?

There is a general mechanism to get direct forms:

urn:x-hp-direct-predicate:http_//www.w3.org/2000/01/rdf-schema#subClassOf

It can be used via a prefix.

Often it is done by having the inference and base data graphs available in the same dataset so it is a matter of using GRAPH.

[VladimirAlexiev]

Often it is done by having the inference and base data graphs available

I described this as a Workaround above (Q2). Let me clarify: given these triples

:x :p :y.
:p rdfs:domain :a.
:a rdfs:subClassOf :b.

An RDFS inferencer will infer:

:x rdf:type :a. # domain
:x rdf:type :b. # subclass
:a rdfs:subClassOf :a. # reflexivity
:b rdfs:subClassOf :b. # reflexivity

But I'd like only these as "direct":

:x sesame:directType :a. # domain
:a sesame:directSubClassOf :b. # explicit