OWL ontology: SPARQL query a range or domain of an ObjectProperty when they're unionOf classes - sparql

Question

I have to query an ontology version 1.4, with a SPARQL 1.1 endpoint, so I can't use the OWL 2 semantic like ClassAssertion etc...
several properties in the ontology have this format:
<owl:ObjectProperty rdf:about="&km4c;hasGeometry">
<rdfs:comment>some services and all railway elements have a specific geometry like polygons or linestrings</rdfs:comment>
<rdfs:range rdf:resource="&gis;Geometry"/>
<rdfs:domain>
<owl:Class>
<owl:unionOf rdf:parseType="Collection">
<rdf:Description rdf:about="&km4c;RailwayElement"/>
<rdf:Description rdf:about="&km4c;Service"/>
</owl:unionOf>
</owl:Class>
</rdfs:domain>
</owl:ObjectProperty>
with domain or range that is a Union of more than one class. The problem is that i want to retrieve all the classes with a certain domain and range, but with the following query:
SELECT DISTINCT ?p
{
?p rdfs:range gis:Geometry.
?p rdfs:domain km4c:Service
}
i get no result, instead of km4c:hasGeometry.
Is there a way to somehow look inside the Collection with this kind of goal?

Answer 1

First, note that the semantics of union of domains and ranges may not be what you expect. In OWL, when you say that class D is a domain of property P, it means that whenever you have an assertion P(x,y), you can infer that D(x). That means that if a domain of P is a union C ⊔ D, then from P(x,y), you can infer that x is an element of C ⊔ D; i.e., that x is either a C or a D, but you don't necessarily know which. For instance, you might define:
hasWings rdfs:domain (Airplane ⊔ Bird)
Then, from hasWings(x,2), you could infer that x is an Airplane or a Bird, but you still wouldn't know which.
Anyhow, if you still want a union class as a domain, you can do that. In the RDF serialization of the mapping of the OWL ontology, the unioned classes are in an RDF list. It's a little bit more complicated to query over those, but you can certainly do it. Since you didn't provide a complete OWL ontology, we can't query over your actual data (in the future, please provide complete, minimal working data that we can use), but we can create a simple ontology. There are two classes, A and B, and two properties, p and q. The domain of p is A, and the domain of q is A or B:
<rdf:RDF
xmlns:rdf="http://www.w3.org/1999/02/22-rdf-syntax-ns#"
xmlns="http://example.org/"
xmlns:owl="http://www.w3.org/2002/07/owl#"
xmlns:rdfs="http://www.w3.org/2000/01/rdf-schema#"
xmlns:xsd="http://www.w3.org/2001/XMLSchema#">
<owl:Ontology rdf:about="http://example.org/"/>
<owl:Class rdf:about="http://example.org/#A"/>
<owl:Class rdf:about="http://example.org/#B"/>
<owl:ObjectProperty rdf:about="http://example.org/#q">
<rdfs:domain>
<owl:Class>
<owl:unionOf rdf:parseType="Collection">
<owl:Class rdf:about="http://example.org/#A"/>
<owl:Class rdf:about="http://example.org/#B"/>
</owl:unionOf>
</owl:Class>
</rdfs:domain>
</owl:ObjectProperty>
<owl:ObjectProperty rdf:about="http://example.org/#p">
<rdfs:domain rdf:resource="http://example.org/#A"/>
</owl:ObjectProperty>
</rdf:RDF>
SPARQL syntax is much more like the N3/Turtle serialization of RDF, so it's helpful to see that serialization, too. The unionOf list is much clearer here:
#prefix : <http://example.org/> .
#prefix rdf: <http://www.w3.org/1999/02/22-rdf-syntax-ns#> .
#prefix owl: <http://www.w3.org/2002/07/owl#> .
#prefix xsd: <http://www.w3.org/2001/XMLSchema#> .
#prefix rdfs: <http://www.w3.org/2000/01/rdf-schema#> .
<http://example.org/#A>
a owl:Class .
<http://example.org/#p>
a owl:ObjectProperty ;
rdfs:domain <http://example.org/#A> .
<http://example.org/#B>
a owl:Class .
<http://example.org/#q>
a owl:ObjectProperty ;
rdfs:domain [ a owl:Class ;
owl:unionOf ( <http://example.org/#A> <http://example.org/#B> )
] .
: a owl:Ontology .
Now you can use a query like this to find properties and their domains, or the unioned classes if one of the domains is a union class:
prefix : <http://example.org/>
prefix rdf: <http://www.w3.org/1999/02/22-rdf-syntax-ns#>
prefix owl: <http://www.w3.org/2002/07/owl#>
prefix xsd: <http://www.w3.org/2001/XMLSchema#>
prefix rdfs: <http://www.w3.org/2000/01/rdf-schema#>
select ?p ?d where {
?p rdfs:domain/(owl:unionOf/rdf:rest*/rdf:first)* ?d
filter isIri(?d)
}
-----------------------------------------------------
| p | d |
=====================================================
| <http://example.org/#q> | <http://example.org/#A> |
| <http://example.org/#q> | <http://example.org/#B> |
| <http://example.org/#p> | <http://example.org/#A> |
-----------------------------------------------------
The interesting parts of that query are:
?p rdfs:domain/(owl:unionOf/rdf:rest*/rdf:first)* ?d
which says you follow a path from ?p to ?d, and the path:
starts with rdfs:domain
and is followed by zero or more repetitions of:
owl:unionOf
followed by zero or more rdf:rest
followed by a single rdf:first
It's not exactly related to this question, but you may find the discussion of querying RDF lists in this answer (disclosure: my answer) to Is it possible to get the position of an element in an RDF Collection in SPARQL? useful.
Then, I also added
filter isIri(?d)
because otherwise we get the node that represents the union class, but that's a blank node that you (probably) don't want.