1

The domain and range of owl:equivalentClass raises some questions to which I have been unable to locate the answers.

First of all, in OWL itself, the domain and range of owl:equivalentClass is set to rdfs:Class. This domain and range seem to clash somewhat with the definition given in the spec (http://www.w3.org/TR/owl-ref/#equivalentClass-def), where it mentions that:

"The meaning of such a class axiom [i.e. owl:equivalentClass] is that the two class descriptions involved have the same class extension (i.e., both class extensions contain exactly the same set of individuals)."

This suggests a more specific domain and range of owl:Class instead of rdfs:Class (considering individuals is mentioned rather than instances).

Using Jena's reasoner engine, I've noticed that the OWL inference rules contain the following interesting elements: (1) Axioms stating owl:equivalentClass has the domain and range owl:Class (2) The rule that two resources that are rdfs:subClassOf each other must be owl:equivalentClass to each other.

So, basically, here the domain and range is restricted to owl:Class, which seems understandable. If its domain and range is narrowed down like that, though, shouldn't the rule at (2) be confined as well to state that both resources should be of type owl:Class in order to entail the owl:equivalentClass relation? After all, the rdfs:subClassOf relation can work perfectly well on rdfs:Classes. If the classes that are subclasses of each other are rdfs:Classes, then surely - if the definition of owl:equivalentClass has been narrowed down in (1) - the rule should not immediately imply that the rdfs:Classes are actually owl:Classes.

Does anyone know the answer as to what the correct interpretation is of owl:equivalentClass, possibly providing an authoritative source?

Thanks in advance!

Semper
  • 11
  • 1
  • StackOverflow is a site about practical programming questions, and your question is theoretical. I see that you already posted it to http://answers.semanticweb.com/ which is a much better place indeed – JimiDini Oct 31 '14 at 09:19
  • 1
    @JimiDini This is definitely a good fit for answers.semanticweb.com, but I think this is probably a good fit for Stack Overflow too. It's a specific technical question about a language standard, and Stack Overflow has plenty of [tag:language-lawyer] questions. – Joshua Taylor Oct 31 '14 at 12:42
  • Crossposted at http://answers.semanticweb.com/questions/30864/is-the-domain-range-of-owlequivalentclass-rdfsclass-or-owlclass – Joshua Taylor Oct 31 '14 at 12:42

1 Answers1

2

The relevant document is OWL 2 Web Ontology Language RDF-Based Semantics (Second Edition). Section 6.4 gives axiomatic triples for the RDF-Based Semantics, and it includes triples that specify the domain and range of owl:equivalentClass as rdfs:Class:

6.4 A Set of Axiomatic Triples

Table 6.2: Axiomatic Triples for the Properties of the OWL 2 RDF-Based Vocabulary

owl:equivalentClass rdf:type rdf:Property .
owl:equivalentClass rdfs:domain rdfs:Class .
owl:equivalentClass rdfs:range rdfs:Class .

Note that OWL also has a direct semantics which doesn't make any use of RDF at all. It defines the semantics in terms of sets of individuals (for classes) and sets of pairs of individuals (for properties). In that case, there'd be no concept of owl:equivalentClass being a property with an rdfs:domain and rdfs:range at all.

As to reasoning with Jena, Jena's reasoners are designed around OWL1, not OWL2, so if there's a significant difference in this part of the spec, Jena would probably be closer to the OWL1 spec. Jena's reasoners are also, as far as I recall, more designed toward OWL Full than OWL DL, and might not care so much about preserving any difference between rdfs:Class and owl:Class.

Joshua Taylor
  • 84,998
  • 9
  • 154
  • 353
  • 1
    There is a long related email and thread here (OWL 1 http://lists.w3.org/Archives/Public/www-rdf-comments/2003JulSep/0332.html ) – Ignazio Nov 01 '14 at 20:36
  • Basically, the axioms may use rdfs:Class, but the classes that can be referenced meaningfully (i.e., no syntax errors and no OWL Full only) are those which happen to be OWL classes as well. So, to me it looks like the situation is very close to the axioms actually using owl:Class. – Ignazio Nov 01 '14 at 20:39