On 27/12/10 17:11, Benson Margulies wrote:
Given a TDB with a substantial amount of content, and a desire to do a
bit of inference, I'm wondering about how I'd write SPARQL to do my
own inference.
Two examples:
Say that I defined some things like:
<owl:ObjectProperty rdf:about="&rex;corefLink">
<rdfs:subPropertyOf rdf:resource="&owl;sameAs"/>
</owl:ObjectProperty>
How would I write sparql for 'sameAs and all of its subprops?'
Subclass is:
?s rdf:type/rdfs:subClassOf* ?T
SubProperty is less pretty:
?s ?p ?o .
?p rdfs:subPropertyOf* ?P
OWL:sameAs
?entity (owl:sameAs|^owl:sameAs)* ?s
This last one needs to be used with care.
Of course, I'm violating a prime directive and perhaps prematurely
optimizing here, feel free to tell me to just fire up the inference
system and see what happens.
It is sometimes better to do this during data preparation, not at query time
See also riot.infer -- this isn't fully integrated into the TDB loader
yet but ...
infer --rdfs=MyVocab.rdfs ... data files ... | tdbloader -- -
Firing up the inference system is also a good idea - and gives access to
a lot more inference capabilities. It is limited to in-memory use.
SPARQL Update could be used to materialize inferences (slowly?) as well:
INSERT { ?s rdf:type ?T } WHERE { ?s rdf:type/rdfs:subClassOf+ ?T }
Note the change from * to +
INSERT { ?s ?P ?o } WHERE { ?p rdfs:subPropertyOf+ ?P . ?s ?p ?o }
Andy
