W3C home > Mailing lists > Public > public-owl-wg@w3.org > August 2008

Re: ISSUE 131 (OWL R Unification): Different semantics on syntactic fragment

From: Ian Horrocks <ian.horrocks@comlab.ox.ac.uk>
Date: Thu, 14 Aug 2008 22:46:58 +0100
Message-Id: <888FDDC9-C721-420F-9620-D64078367BD2@comlab.ox.ac.uk>
Cc: "Ivan Herman" <ivan@w3.org>, <public-owl-wg@w3.org>, "Alan Wu" <alan.wu@oracle.com>
To: Michael Schneider <schneid@fzi.de>

On 13 Aug 2008, at 10:10, Michael Schneider wrote:

> Hi Ian!
>
> Ian Horrocks wrote:
>
>> However, my hope and expectation is that conformance for OWL RL will
>> be defined such that a reasoner is conformant if it is sound w.r.t.
>> the OWL Full/RDF semantics, and is complete w.r.t. the entailments
>> derived using the rule set.
>
> This statement starts to clarify things for me now.
>
> So soundness and completeness wouldn't be, as usual, specified  
> w.r.t. a
> *single* semantics, but there would be *two*: The ruleset for defining
> completeness (the "lower bound" semantics), and OWL 2 Full  
> semantics for
> soundness (the "upper bound" semantics).

I was talking about specifying *conformance*, not soundness and  
completeness per se. Given that an OWL RL reasoner accepts all OWL  
Full ontologies (aka RDF graphs), it makes sense (to me at least)  
that conforming reasoners should at least be sound for them -- this  
seems essential for interoperability. It would clearly be  ridiculous  
to demand that conformant implementations also be complete, and the  
question thus arises as to what "minimum" condition, if any, to set  
on conformance. One possibility is to be at least as complete as a  
rule based implementation would be.

>
> This would, of course, mean that there can be two different  
> reasoners, which
> happen to produce different sets of inferences for the same  
> ontology, but
> still, they can both call themselves be "compliant" rasoners. The  
> only thing
> which can be said in such a case is that the two sets of produced  
> inferences
> are both upper sets of the inferences expected from the ruleset. I  
> wonder
> whether this can lead to interop problems; others will have to tell  
> me.

It's not perfect for interoperability, but it seems to me that  
introducing a 3rd semantics that is highly incompatible with the  
existing ones would be much worse. Indicative of this is the fact  
that the 3rd semantics proposal would oblige us to have a way of  
stating the intended semantics of an ontology. The result would be  
that for a large class of ontologies with "OWL R" as the intended  
semantics, there would be literally zero interoperability --  
reasoners for OWL Full, DL, or any other profile would be unsound (in  
general) for these ontologies.

In contrast, in the "merging" proposal, we wouldn't need intended  
semantics. The only drawback I see from an interoperability POV is  
that for a very similar class of ontologies, OWL R reasoners couldn't  
guarantee to be complete. However, as this class is syntactically  
identified, tools can (optionally) warn users when this problem might  
arise.


> The question remains whether this completeness definition (w.r.t. the
> ruleset) will hold for all RDF graphs, or is it restricted to  
> ontologies
> from the syntactic fragment only? If it holds for all RDF graphs,  
> this would
> mean that one cannot simply use a classic OWL DL reasoner for complete
> reasoning in OWL RL. Instead, one has to make sure that such a  
> reasoner is
> also able to produce the necessary inferences for all of RDF.

Actually, I much prefer your idea of defining conformance w.r.t. the  
syntactic fragment, i.e., conformant OWL RL reasoners must be  
complete for query answering as defined in the current section 4.4.  
This would be a slightly less strict condition, but it has the  
advantage of being non-procedural and of coinciding with the  
syntactically defined class of ontologies for which rule-based  
implementations are complete.


>
> Technically, this would not be too hard to achieve: Just add a ruleset
> reasoner, and delegate to it every RDF graph, which the syntax checker
> detects to be not from the syntactic fragment of OWL RL (or at  
> least not
> valid OWL DL).
>
>> A reasoner of the kind you describe is
>> trivially complete for entailments derived using the rule set, and
>> also is trivially sound w.r.t. the OWL Full/RDF semantics;
>
> Yes.
>
>> it would therefore be a conformant (compliant if you prefer) OWL RL
> reasoner.
>
> Ok.
>
> So to summarize, I still have two questions:
>
>   * Does an OWL RL compliant reasoner need to produce the ruleset  
> inferences
> for every RDF graph, or only for ontologies from the syntactic  
> fragment?
> (Yes, I know, that's a conformance question. :))

This is something we can discuss/decide. I didn't think of it before,  
but I much prefer the latter solution for the reasons I mentioned above.

>
>   * Is it possible that this approach will lead to interop problems?

This is something else we can discuss. It seems to me to be much less  
bad than the alternative.

Regards,
Ian


>
> Cheers,
> Michael
>
> --
> Dipl.-Inform. Michael Schneider
> FZI Forschungszentrum Informatik Karlsruhe
> Abtl. Information Process Engineering (IPE)
> Tel  : +49-721-9654-726
> Fax  : +49-721-9654-727
> Email: Michael.Schneider@fzi.de
> Web  : http://www.fzi.de/ipe/eng/mitarbeiter.php?id=555
>
> FZI Forschungszentrum Informatik an der Universität Karlsruhe
> Haid-und-Neu-Str. 10-14, D-76131 Karlsruhe
> Tel.: +49-721-9654-0, Fax: +49-721-9654-959
> Stiftung des bürgerlichen Rechts
> Az: 14-0563.1 Regierungspräsidium Karlsruhe
> Vorstand: Rüdiger Dillmann, Michael Flor, Jivka Ovtcharova, Rudi  
> Studer
> Vorsitzender des Kuratoriums: Ministerialdirigent Günther Leßnerkraus
>
Received on Thursday, 14 August 2008 21:47:40 GMT

This archive was generated by hypermail 2.2.0+W3C-0.50 : Thursday, 14 August 2008 21:47:43 GMT