0

In OWL/DL, it is possible to have a property/role that is a Sub[Object/Data]PropertyOf/subrole of more than one property/role, but I never found a motivating example in the W3C OWL documentation or in DL papers.

Do you know of any published OWL ontologies or DL paper that includes such a motivating example that I have missed (examples of your own are welcome in comments) ?

Rhangaun
  • 1,430
  • 2
  • 15
  • 34

1 Answers1

3

This might be too broad of a question, but it's easy to think of properties that have multiple superproperties.

Logical motivation

After all, that p is a subproperty of q just means that "x p y" implies "x q y". So all you need is some kind of relationship that simultaneously implies two different kinds of relationships:

hasSon ⊑ hasMaleRelative, hasDescendant

This makes sense, since having a son X implies having X as a male relative, and having X as a descendant, but it's not the case that either

hasMaleRelative ⊑ hasDescendant

or

hasDescendant ⊑ hasMaleRelative

That might not be a particularly motivating example, but it's simple and clear. Any time you a single property implies at least two others, you've got a property with multiple super properties.

The trivial case of multiple subproperties

Note also that because subPropertyOf is transitive, as soon as you have a multi-level hierarchy, you have multiple super properties. E.g., 

p 
|
+- q
   |
   +- r

r has both q and p as super-properties.

Interoperability and Ontology Matching

One important reason for declaring multiple superproperties of a property is in matching different ontologies. For instance, if you're defining an ontology where text can be associated with some resource, you might want to relate your ontology with both the RDFS vocabulary and with Dublin Core. As a result, you might end up with

ex:myNote rdfs:subPropertyOf rdfs:comment, dc:description

Joshua Taylor
  • 84,998
  • 9
  • 154
  • 353
  • I think you offer a good example (I should have explicitly excluded the trivial case of a multi-level hierarchy). – Rhangaun Feb 17 '15 at 18:56
  • @Skeptic Another important case (I've updated my answer) is when you're relating multiple ontologies. – Joshua Taylor Feb 17 '15 at 20:55
  • Interroperability is handled using equivalentProperty according to the examples I've seen. – Rhangaun Feb 18 '15 at 16:00
  • 1
    @Skeptic I'd bet that a bunch of those cases aren't as semantically correct as their authors would hope. As soon as you get an "equivalentProperty" relation, you get implications in both directions. I know that dbpedia-owl:abstract relates DBpedia resources with their abstracts, and I'd be confident making it a subproperty of rdfs:comment, but it certainly wouldn't be an *equivalent property*. Not everything that's a comment is an abstract. It's *especially* important when properties have domains and ranges. If the domain of dbpedia-owl:abstract is, e.g., dbpedia-owl:Article or... – Joshua Taylor Feb 18 '15 at 16:26
  • @Skeptic ...something like that, then if I make rdfs:comment an *equivalent* property, then any time I have `x rdfs:comment y`, then I get `x dbpedia-owl:abstract y` and then I get `x rdf:type dbpedia-owl:Article`. There are times when owl:equivalentProperty is the right choice, but usually it's stronger than what should be used, and if a reasoner is present, it's will often lead to unintended results. – Joshua Taylor Feb 18 '15 at 16:28