Re: Final text for Basic Graph Patterns

On 18 Jan 2006, at 23:12, Pat Hayes wrote:
>>> SELECT ?x WHERE {:a rdf:type ?x .}
>>
>> Look, this is not a legal OWL-DL query.
>
> This query pattern has legal OWL-DL instances, so why is it not a  
> legal OWL-DL query?

Look, we already agreed that we are not here to do research.
First of all, I have been sloppy in my statement, since OWL-DL  
queries do not exist - we are defining them now here.
Now, there is *no* theoretical nor practical result that even  
considers queries of that kind in the DL literature. Queries that  
have been studied in the world of DL are only first order. Full stop.

>> We pointed out in our old proposed text (and also in the new one).  
>> We allow in OWL-DL only queries that carry similar syntactic  
>> restrictions as OWL-DL expressions.
>
> Saying that a thing is in a class, using rdf:type, can be a legal  
> OWL-DL expression. Many OWL ontologies consist of little else. -DL  
> does not allow restrictions and other definitions to use rdf:type,  
> but it does not prohibit simple assertions of membership in an OWL  
> class.

Yes, but in no DL statements with existential variables in place of  
classes have ever been considered.

>> This boils down to allow only for non-high order RDF graphs as  
>> BGPs (defined in [1]). So the above query is not of our concern in  
>> OWL-DL.
>
> Well, of course you are free to define things your way, but that  
> does not seem to alter the OWL DL specs, which do not mention  
> "higher-order" graphs.

I have used this definition to say that existential variables in  
place of concepts, or variables in place of predicates, have never  
been considered in DLs.

>> As far as OWL-full is concerned, my very personal position is that  
>> I don't care, since I don't believe it makes *currently* any sense  
>> at all :-)
>
> Maybe you should read some more recent logic texts :-). None of the  
> OWLs are higher order.

I am not saying that, but it is not relevant.

> But let us ignore Full for now, I agree.

Thanks God.

>> So, whatever you say about OWL-full, is fine for me...
>
> I wasn't meaning to go into OWL Full. The entailments I mentioned  
> are all OWL-DL valid.

No, because you have to point me a definition in some DL paper where  
higher order queries are mentioned, their semantics well defined,  
their computational properties analysed, and effective sound and  
complete algorithm proposed, and by the way including the OWL-DL  
dialect. You are my guest.

It is so true that you and Bijan are wondering how to define the  
meaning of such query.

>>> Now, this dataset OWL-entails the *existence* of a triple  
>>> intersection and three double intersections, all with :a in them.  
>>> So are these reasonable answer bindings for such a query? I see  
>>> no good reason why they should not be: in fact, one could  
>>> reasonably take the position that the *only* non-redundant answer  
>>> here would be to bind ?x to a term representing the most  
>>> restrictive intersection. But to construct such a term would  
>>> require the use of the RDF collection vocabulary when presenting  
>>> the answer.
>>
>> Even if we allow high-order queries like the above in OWL-DL, the  
>> scoping set B would be restricted to be URIs in OWL-DL, so the  
>> problem doesn't show up in OWL-DL anyway!
>
> We could so restrict things, of course: but my point was that such  
> a restriction might not be semantically/pragmatically correct in  
> the OWL case.

In the OWL-DL case it is the only one that makes sense given the  
current state of the research and practice. The DL people will be  
more than happy with that; they were just afraid that SPARQL would  
have been semantic-less and that it wouldn't scale to the current  
theories and practices of DLs. With our proposal the community is for  
sure more than happy, I can make a poll, if you want :-)

> I agree some form, sufficiently restricted, of OWL querying can be  
> made to work in the SPARQL framework, but if we try to design a  
> querying standard which is really based correctly on actual OWL-DL  
> entailment, and not on a narrow restriction of it, then I think we  
> will need to reconsider these definitions. But perhaps we all agree  
> on that point.

What you call "a narrow restriction of it" is the only one existing  
in DLs, and the only one we should bother about. And I am the DL  
expert here :-)

--e.

Received on Thursday, 19 January 2006 00:33:32 UTC