Hi all,
I have some ontologies (Proton modules [1]) in a triplestore (3store [2]), and using SPARQL, I'd like to find all properties that can exist between 2 instances - that are also in the store. So, what I need to do basically is - am I right btw ?: 1) find types of these instances (direct and inferred ones - as the store do inference for instances type); 2) find all properties that have these types as domain/range 3) return all subproperties of the previous one (as I do not infer suproperty inheritance in the store, so some properties are not retrieved by 2) The following query works fine: PREFIX xsd: <http://www.w3.org/2001/XMLSchema#> PREFIX rdfs: <http://www.w3.org/2000/01/rdf-schema#> PREFIX rdf: <http://www.w3.org/1999/02/22-rdf-syntax-ns#> SELECT DISTINCT ?property WHERE { // 1) <http://foobar#instance1> rdf:type ?domain . <http://foobar#instance2> rdf:type ?range . // 2) _:p rdfs:domain ?domain . _:p rdfs:range ?range . // 3) ?property rdfs:subPropertyOf _:p . } (+ Union to get domain / range inverse) Yet, in the original ontology, some subproperties got an additionnal domain / range that conflict with the type of instances. Eg, I got this in my ontology: Person rdfs:subClassOf Agent Group rdfs:subClassOf Agent Company rdfs:subClassOf Group hasMember rdfs:domain Group hasMember rdfs:range Agent hasEmployee rdfs:subPropertyOf hasMember hasEmployee rdfs:range Person So the query before applied to Company instances will return hasEmployee which _shouldn't_ be applied between 2 companies (This is what is done in Protege where hasEmployee exists only between Group and Person, but using RDFS inference rules, it should exist ?. Maybe I missed something here ? What I want to do to avoid this problem is to remove from query results the properties that got a domain / range which is not a superclass (or the class itself) of the direct type of my instances. First, is that the good way to do ? I can't see any other solution for this at the moment. To do so, my query must: 1) find direct type of these instances 2) find all types of these instances 3) find all properties that have these types as domain/range 4) find all subproperties of the previous one 5) remove subproperties where domain or range is not a parent of "1" I make a query that works until 4, but I can see how to remove properties I don't want, because of this: PREFIX xsd: <http://www.w3.org/2001/XMLSchema#> PREFIX rdfs: <http://www.w3.org/2000/01/rdf-schema#> PREFIX rdf: <http://www.w3.org/1999/02/22-rdf-syntax-ns#> PREFIX direct: <http://triplestore.aktors.org/direct/#> SELECT DISTINCT ?p ?domain ?range WHERE { // 1) <http://foobar#instance1> direct:type _:direct_subject_type . <http://foobar#instance2> direct:type _:direct_object_type . // 2) <http://foobar#instance1> rdf:type _:subject_type . <http://foobar#instance2> rdf:type _:subject_type . // 3) _:property rdfs:domain _:subject_type . _:property rdfs:range _:object_type . // 4) ?p rdfs:subPropertyOf _:property . // 5) ?p rdfs:subPropertyOf ?z . ?z rdfs:domain ?domain . ?z rdfs:range ?range . } Then I should test ?domain / ?range regarding 5), but the problem is that I got properties that both conflict and are OK: }, { "p": {"value": "http://proton.semanticweb.org/2005/04/protont#subRegionOf";, "type": "uri" }, "domain": {"value": "http://proton.semanticweb.org/2005/04/protons#Entity";, "type": "uri" }, "range": {"value": "http://proton.semanticweb.org/2005/04/protont#Location";, "type": "uri" } }, { "p": {"value": "http://proton.semanticweb.org/2005/04/protont#subRegionOf";, "type": "uri" }, "domain": {"value": "http://proton.semanticweb.org/2005/04/protons#Entity";, "type": "uri" }, "range": {"value": "http://proton.semanticweb.org/2005/04/protons#Entity";, "type": "uri" } }, { subRegionOf should be removed as Location is not a Parent of "Company" (1 result), yet, Entity is (2nd), so even when adding the subClass condition between original type and properties, I'll keep the 2nd result, which I don't want as the 1st is false. Any idea to solve this ? Or maybe another way to find all properties that can exist between 2 instances ? And thanks for those that read this mail totally :) Best, Alex. [1] http://proton.semanticweb.org [2] http://threestore.sf.net
