rdfs:subPropertyOf 和 rdfs:domain

Question

Is the following statement possible?

:a rdfs:subPropertyOf :b
:b rdfs:domain :aDomain

Thus, :a rdfs:domain :aDomain

Answer 1

Are these two statements possible?

:a rdfs:subPropertyOf :b .
:b rdfs:domain :aDomain .

Certainly. The first states that every :a of something is also :b , and the seconds states that every object of :b is an :aDomain .

Is this statement possible?

:a rdfs:domain :aDomain .

Yes, it states that every object of :a is an :aDomain .

Is it implied by the first two statements, meaning that adding this statement to a graph already containing the two does not change whatever can be inferred about the known individuals? Yes. In the first case, being an object of :a implies also being an object of :b , and thus being an :aDomain , which implies that an object of :a is an :aDomain , exactly what the statement says.

Answer 2

TL;DR: according to the RDFS semantics, if everything that has property :a also is of type :aDomain , it does not automatically imply that :a rdfs:domain:aDomain ; according to the OWL semantics, if everything that has property :a also is of type :aDomain , it logically implies that :a rdfs:domain:aDomain .

The answer to this question is more subtle than it may seem because there is a slight difference between the semantics of RDFS and that of OWL. Stanislav Kralin highlight this subtlety in his laconic comment to the question. The RDF snippet given by Costas can be interpreted according to the RDFS semantics or the OWL semantics (or even the RDF semantics , or the simple semantics , or other semantic extensions ).

According to simple semantics and RDF semantics, the terms rdfs:subPropertyOf and rdfs:domain do not have any special meaning and therefore the conclusion :a rdfs:domain:aDomain does not hold.

According to RDFS semantics, the conclusion does not hold either, but it is less obvious. In RDFS, if you have:

:a rdfs:subPropertyOf :b .
:b rdfs:domain :aDomain .
?x :a ?y .

for any IRI or blank node ?x and any IRI or blank node or literal ?y , then you can conclude that:

?x rdf:type :aDomain .

but you cannot conclude that:

:a rdfs:domain :aDomain .

This is because RDFS does not have an extensional semantics . This means that if everything that has the type C also has the type D , you cannot conclude (from this alone) that C is a subclass of B , and if everything that has the property p has type C , then you cannot conclude (from this alone) that p has the domain C . In 2004, RDF 1.0 presented an alternative version of RDFS with extensional semantics , that highlighted the issue. In 2013, Franconi et al. discussed The Logic of Extensional RDFS and its properties (link also provided by Stanislas Kralin in his comment to the OP).

Finally, OWL semantics (both the Direct Semantics or the RDF-based semantics ) has an extensional semantics, so the conclusion holds when interpreting the RDF snippet according to these semantics:

:a rdfs:subPropertyOf :b .
:b rdfs:domain :aDomain .

entails (that is, logically implies):

:a rdfs:domain :aDomain .