0

I have OWL ontology, Protege 5. I want starting form a source class S, to retrieve all destination classes D where S is related to Ds by means of to-one cardinality through objectProperty i.e S ([0..*] - 1) D despite what is the connecting objectProperty.

In DL, it would be something similar to :

S ⊑ =1r:D;

where S is my source concept, D is a variable concept and r is a variable objectProperty.

First Point: How to express this in Manchester Syntax to use it in Protege DL query?

Second Point: For such query, what are the considered types of property restriction that the reasoner will consider when trying to answer the query? e.g cardinality restriction, functional properties, someValuesFrom, allValuesFrom?

Thanks.

Median Hilal
  • 1,483
  • 9
  • 17
  • I posted an answer, but realized that I may have misunderstood the scope of your variables a bit. Are you asking for S and D such that each s in S, for each object property p, is related to exactly one d in D? Or are you asking for S and D such that each s in S is related to exactly one D, period. E.g., you could have (=1 p.D) and (=1 q.D), but then some s might be such that p(s,d1) and q(s,d2). Then s would be related to two D's. Is that OK? I think you can only get the first, because owl:topObjectProperty relates *every* pair of individuals, so in OWL every pair is connected. – Joshua Taylor Oct 13 '14 at 14:33
  • Actually, I'm only considering TBOX, I have no individuals in my OWL. What I exactly want is: having a class S at hand (constant), I want to find all the classes Class1, Class2, Class3... that are S ⊑ =1p :Class1; S ⊑ =1q:Class2; S ⊑ =1r:Class3... where S is defined as the domain for each of p,q,r and Class1,Class2,Class3 are their ranges respectively. In other words, at the beginning I only have one class S at hand and I want to find all the classes Di related to S by an objectProperty pi where S is the property domain, Di is property range and S subClassOf pi exactly 1 D. – Median Hilal Oct 13 '14 at 16:46
  • Having (=1 p.D) and (=1 q.D) is OK. – Median Hilal Oct 13 '14 at 16:49

1 Answers1

1

You can't write the kind of query that you're asking about in the DL query syntax. The DL query syntax doesn't have any place for variables; all you can write are class expressions, after which you can ask for the individuals in that class, or subclasses or superclasses of the class. So the answer to the first question is that you can't express your query in the Manchester syntax.

For the second point, there may be reasoners that can help you draw a conclusion here. For instance, you might be able to use a reasoner that supports SPARQL queries and write the template of the class expression with variables. You'd end up with something like this (but this is untested):

prefix xsd:  <http://www.w3.org/2001/XMLSchema#>
prefix rdfs: <http://www.w3.org/2000/01/rdf-schema#>
prefix owl:  <http://www.w3.org/2002/07/owl#>

select ?s ?p ?d where {
  ?s rdfs:subClassOf [ a owl:Restriction ; 
                       owl:qualifiedCardinality "1"^^xsd:nonNegativeInteger ;
                       owl:onProperty ?p ;
                       owl:onClass ?d ]
}

I don't know whether or not most OWL reasoners will be able to handle that though. As mentioned above, the typical task for an OWL reasoner is to look at class expression and determine its subclasses, and superclasses. This is getting more complicated.

Joshua Taylor
  • 84,998
  • 9
  • 154
  • 353