Re: RIF must cover RDF triples as data

Dan Connolly wrote:
>>  [...]
>>(How) is this different from
>>
>>CONSTRUCT {?s ?q ?o}
>>WHERE { ?p rdfs:subPropertyOf ?q. ?s ?p ?o }
>>
>>?
> 
> No, I don't mean for it to be different at all.
> That's exactly the point... by my reading of the current RIF
> semantics proposal[1], we need a holds() predicate in order
> to match the semantics of SPARQL (and RDFS and OWL Full).
> 
> [1] http://www.w3.org/2005/rules/wg/wiki/A.1_Basis%3A_Positive_Conditions
> last edited 2006-09-19 16:10:24
> 
> 
>>... I mean, I know that it is different, but my problem is more the 
>>following:
>>
>>In what *context* is the inferred triple true? What is the context of 
>>the body triples what is the context of head triples here?
> 
> 
> Hmm... I don't understand this question. I don't see "context"
> in [1]. Could you perhaps phrase the question some other way?

I belief that there should be a context...

what I mean is:
The result graph of a SPARQL CONSTRUCT graph and the original dataset 
are clearly separate, since the CONSTRUCTed triples hold in the result 
graph, not in the dataset, however, in the rule:

  holds(rdfs:subPropertyOf, ?p, ?q) /\ holds(?p, ?s ?o)
       => holds(?q, ?s ?o)

there is no such separation, if I'd see this rule together with some rdf 
triples in the same document/data(+rule)set, I'd assume that the 
inferred triples also "belong" to the same document/context.


best,
axel


-- 
Dr. Axel Polleres
email: axel@polleres.net  url: http://www.polleres.net/

Received on Wednesday, 11 October 2006 15:04:46 UTC