Re: subgraph/entailment

On Sep 7, 2005, at 10:35 AM, Seaborne, Andy wrote:

> Bijan Parsia wrote:
[snip]
>> Plain RDF entailment may be uninteresting, but I think Enrico's point 
>> is well taken.
>> I tried this query:
>> 	PREFIX rdf: <http://www.w3.org/1999/02/22-rdf-syntax-ns#>
>> 	SELECT ?p
>> 	WHERE {  ?p rdf:type rdf:Property }
>> in http://sparql.org/query.html and http://librdf.org/query and got 
>> no hits. I think I could have reasonably expected to have gotten all 
>> the asserted properties in the document (by entailment rule rdf1).
>
> Which model did you query?  Did you read the service description as to 
> what the contract on that service with that model is?  Where does it 
> say it gives simple/RDF/RDFS entailment?

I'm not saying it was a *right* expecation, but a reasonable one.

>> I would have perhaps been surprised to have gotten all the properites 
>> in the axiomatic triples, but probably would have concluded that that 
>> was correct. As an implementor, I would have thought, reading the 
>> specs, that both the above implementions got it wrong because my 
>> understanding of a proper RDF graph *for the purpose of query* would 
>> have including the entailments forced by the semantics.
>
> Ah - you are saying that there should never be "zero-entailement".

I'm saying that that's not an unreasonable expectation. I can see 
thinking that "querying an RDF graph" means "querying and RDF under RDF 
entailment".

(I think querying the bare graph is important too. But it's not obvious 
:( I'm going over the relevant RDF documents to try to come up with a 
list of issues.)

> and you also say that you expected some and not others.

Yes. Rather, my naive reaction would be for all the actually mentioned 
properties (i.e., just apply rdf1 without the axiomatic triples). If I 
checked my assumption with the spec, I would have expected rdf1 over 
the graph + axiomatic triples. Thinking of it as a pattern match with 
*no* semantics, I get no answers.

I fear that people will quite naturally expect  "?p rdf:type 
rdf:Property" to simply *mean* "fetch me all the properties mentioned 
in this document". It's very intuitive to read it that way and it's a 
useful query.

> You do need to check the service description to see what the service 
> offers.
> (which is tricky for http://sparql.org/query.html :-)

Yes. And it's not obvious to me from the SPAQRL spec how I determine 
this. Some pointer from the query language would be helpful.

I think that even if for matching purposes, we pretend there is just a 
big ole asserted graph, it's still helpful to distinguish the case 
where the assertions are generated by people or other data generation 
source and those forced into the graph by the semantic conditions and 
entailment relations. It *is* a distinguished subcase. As I think Dan 
said, FROM and friends already go here a little (maybe FROM dataset 
UNDER RDF-semantics???)

Still trying to figure out the minimal good thing.

Cheers,
Bijan.

Received on Wednesday, 7 September 2005 17:38:02 UTC