Avoiding count explosion in querying with SPARQL ontologies containing owl:sameAs triples - sparql

Question

I am trying to build a (local) ontology that describes a finite number of objects, and that links these objects to external resources via the owl:sameAs predicate. However, when I simply query for the number of objects of that kind, I obtain twice as much as the object described. It is clear that also the external resources are counted independently, as what is taken into account is the number of URIs, and not the number of distinct objects.
I have solved this issue in the following way: I assume that the local ontology can be seen as a "reference hub" for knowing basic stuff about these objects, so I select all the objects of a certain kind, and then filter out only those that contain the base URI of the local ontology, i.e.:
# How many objects are there?
PREFIX ch: <http://www.example.com/ontologies/domain#>
SELECT (COUNT(DISTINCT ?elem) AS ?count) WHERE {
?elem a ch:Element.
FILTER (REGEX (STR(?elem) ,"http://www.example.com/ontologies/domain") ).
}
However, I have two concerns with this way of doing:
1) it looks a bit of a hack (even if somehow principled), whilst I would like something that makes more logical sense
2) I have the impression that this query is not very efficient.
I have searched quite a bit here, and on google, but didn't come out with any better solution... any suggestions here?
Thank you very much for any help!

Answer 1

GROUP BY a representative element
If there's some property that should have distinct values for each individual, then you can use it to impose the "equivalent class" structure that you need. E.g., something like this:
prefix ch: <http://www.example.com/ontologies/domain#>
select (count(?label) as ?count) where {
?elem a ch:element ;
rdfs:label ?label .
}
group by ?label
Synthesize a representative element
if there's not a value that will be shared by all elements in an equivalence class, you can still get a representative element from the set by asking for the minimal element in each equivalence class. We can use the IRIs of the elements to order the elements, and use that to select a unique individual. This does presume that each ?elem and all the things that it is the same as have well defined behavior under the str function (and IRIs do).
prefix ch: <http://www.example.com/ontologies/domain#>
select (count(distinct ?elem) as ?count) where {
?elem a ch:element .
filter not exists {
?elem (owl:sameAs|^owl:sameAs)* ?elem_
filter( str(?elem_) < str(?elem) )
}
}