Re: OWL Full proposal (sort of) - addressing my Action

+1  (Michael, I think you've got it!)

On Feb 13, 2008, at 6:20 AM, Ivan Herman wrote:

> Michael,
>
> I do tutorials on RDF & co regularly. When talking about RDFS, and  
> what it does, I *always* use those entailement rules to explain, for  
> example, what subclassing mean more formally. I never ever refer to  
> the model theoretic thing. And that has worked very well in my  
> experience, people get it and they are happy:-)
>
> So if such set of entailement rules were available for OWL Full  
> (even if informative, even if imprecise in some corner cases) that  
> would be really good in my view...
>
> My 2 cents
>
> Thanks
>
> Ivan
>
> Michael Schneider wrote:
>> Hi Ian and Jim!
>> (Ok, this might become a bit controversal, but let's see... :))
>> Maybe there is an option, which might work, and which has  
>> precedence. Ivan came up with an interesting point last week [1].  
>> He stated that the RDF semantics spec [2]     "includes a bunch of  
>> entailment rules which are     the type of axiomatic rules that you  
>> [Jim] refer to". I think this is a relevant observation.
>> As I wrote in [3], the OWL-Full semantics are actually specified as  
>> a set of model-theoretic "semantic conditions". And I believe that  
>> it really is appropriate to /specify/ the OWL-Full semantics for  
>> the different language features in this way. When I have a QCR  
>> described by a set of RDF triples, then defining the meaning of  
>> this collection of triples by set theoretical means is clear and  
>> straightforward, IMO. And I think this also eases the task of  
>> making OWL-Full "downwards" compatible with OWL-DL, since one can  
>> try to "borrow a little bit" from those specifications which  
>> already exist in the DL semantics document. :) Anyway, doing it  
>> this way will give us the desirable rigour.
>> On the other hand, I believe that the semantic conditions often do  
>> not very well show how OWL-Full "works", or how parts of it can be  
>> implemented. Much better is this captured, IMHO, by "triple  
>> entailment rules". By this I mean rules, which get a few RDF  
>> triples as their input, and fire a few new RDF triples as their  
>> output, e.g.:
>>    IF (?x rdfs:subClassOf ?y) AND (?y rdfs:subClassOf ?z)     THEN  
>> (?x rdfs:subClassOf ?z)
>> The "meaning" of a given RDF graph can then be understood to be the  
>> inference graph, which consists of the original triples, and all  
>> those triples which are additionally entailed by all triple  
>> entailment rules.
>> Semantic conditions do not have such a triple-to-triple form, and  
>> for the example above doesn't even exist a semantic condition which  
>> could be regarded to be the "direct model-theoretic analogon".  
>> However, it isn't hard to see that from OWL-Full semantics the  
>> above triple entailment rule can be "derived" (in some sense).
>> The RDF(S) semantics spec [2] contains a section which itemizes a  
>> set of such triple entailment rules [4]. This is an /informative/  
>> section, while the sections which introduce the semantic conditions  
>> are normative. If I had to point someone to a resource where he can  
>> learn how OWL-Full "works", I think I would rather point him to  
>> this collection of triple entailment rules.
>> Regrettably, there is no such collection of triple entailment rules  
>> in the OWL-1.0-Full spec. I would think that having such a  
>> collection would be a win, and I expect the costs for its creation  
>> to be mediocre. Perhaps, such a collection doesn't even have to be  
>> complete, if this is possible at all. But even a "proper selection"  
>> would already be of value, I suppose. Such an informative section  
>> which itemizes triple entailment rules should probably not be part  
>> of the UFDs, since it would still be very technical. Of course, one  
>> could have a few of these rules in a Primer document as examples  
>> for explaining OWL-Full, and then point to the respective section  
>> in the Full semantics spec as a reference. Btw: The above example  
>> of a triple entailment rule does not just accidentally look similar  
>> to a Jena rule [5]. :) AFAIU, the Jena inference support actually  
>> follows the triple entailment rule approach as given in [4], see [6].
>> Cheers,
>> Michael
>> [1] <http://lists.w3.org/Archives/Public/public-owl-wg/2008Feb/0051.html 
>> >
>> [2] <http://www.w3.org/TR/rdf-mt/>
>> [3] <http://lists.w3.org/Archives/Public/public-owl-wg/2008Feb/0068.html 
>> >
>> [4] <http://www.w3.org/TR/rdf-mt/#rules>
>> [5] <http://jena.sourceforge.net/inference/index.html#rules>
>> [6] <http://jena.sourceforge.net/inference/index.html#RDFSintro>
>>> -----Original Message-----
>>> From: public-owl-wg-request@w3.org [mailto:public-owl-wg-request@w3.org 
>>> ] On Behalf Of Ian Horrocks
>>> Sent: Tuesday, February 12, 2008 7:09 PM
>>> To: Jeremy Carroll
>>> Cc: Jim Hendler; Web Ontology Language ((OWL)) Working Group WG
>>> Subject: Re: OWL Full proposal (sort of) - addressing my Action
>>>
>>>
>>> I share Jeremy's desire to maintain a significant degree of  
>>> rigour  w.r.t. OWL Full; in fact I would say the same for all  
>>> species/ fragments/conformance-levels. Moreover, from what Jim  
>>> says in emails  such as [1] it would seem that he has no objection  
>>> to this and  perhaps even believes it to be mandatory. Instead,  
>>> what he seems to  be asking for (in [1] at least) is a more human  
>>> readable explanation  for those who find model theories difficult  
>>> to understand. I don't  see any problem with that -- in fact we  
>>> already have a commitment to  producing such documents and even a  
>>> UFDTF busily engaged in their  development. So, if we were to  
>>> agree to produce a "delta to OWL-1.0- Full" as outlined by Michael  
>>> in [2] plus suitable documentation for  the MT-challenged, would  
>>> this satisfy everyone's requirements?
>>>
>>> Ian
>>>
>>> [1] http://lists.w3.org/Archives/Public/public-owl-wg/2008Feb/0069.html
>>> [2] http://lists.w3.org/Archives/Public/public-owl-wg/2008Feb/0068.html
>> --
>> 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
>
> -- 
>
> Ivan Herman, W3C Semantic Web Activity Lead
> Home: http://www.w3.org/People/Ivan/
> PGP Key: http://www.ivan-herman.net/pgpkey.html
> FOAF: http://www.ivan-herman.net/foaf.rdf
>

"If we knew what we were doing, it wouldn't be called research, would  
it?." - Albert Einstein

Prof James Hendler				http://www.cs.rpi.edu/~hendler
Tetherless World Constellation Chair
Computer Science Dept
Rensselaer Polytechnic Institute, Troy NY 12180

Received on Wednesday, 13 February 2008 14:22:58 UTC