RE: Action-67 some examples on b-nodes issues and their impact on users

Hi, Ivan!

Just a remark on your OWL-Full related question:

Ivan Herman wrote:

>2. RDFS
>
>I would like to understand how exactly the proposed changes affect the 
>semantic compatibility of OWL-Full and RDFS, ie, where would the 
>differences be. My current understanding is that the semantics of OWL 
>Full is a suitable extension of the RDFS semantics, ie, 
>everything that 
>is true in RDFS is also true in OWL Full. 

Yes.

>Will that relationship break?

If bNodes are re-interpreted as skolems in OWL-1.1-Full, yes. In this case, OWL-1.1-Full would not have a truely RDF-compatible model-theoretic semantics anymore. 

>I would expect so, but I would like to have examples on how this would 
>diverge... Jeff, would it be possible to find an example where an RDFS 
>inference would no longer be true in a changed OWL1.1 Full?

Here is an example of an RDF entailment:
 
  { ex:s ex:p _:x } |= { ex:s ex:p _:y }

That's an RDF entailment, because the LHS and the RHS of "|=" have the same meaning under existential bNode semantics:

  ( EXISTS x: ex:s ex:p x ) <==> ( EXISTS y: ex:s ex:p y )

However, with bNodes having skolem semantics, this is not an entailment. The variable name '_:x' differs from the variable name '_:y'. Therefore, these two names *may* denote different things. And therefore, interpretations are possible where the LHS of "|=" is true, while the RHS is false. And this means, that it is a non-entailment.

>A related question is whether it is possible to change the 
>semantics of 
>all this on the DL interpretation only. 

This is the only approach I see. As long as OWL-1.1-Full is intended to be RDF compatible, there's no way to have bNodes as skolems in Full. The only way to let Full have skolem bNodes is to change the RDF semantics spec. But this is not within the scope of this OWL-WG.

The best this OWL-WG can do in this situation is to make sure that there will be no direct usage or reference of existantial bNode semantics in all the semantic conditions, which are given in the OWL-1.1-Full spec. This doesn't, of course, remove existentials from the Full spec, because they come in anyway, since the Full spec extends the RDF(S) spec. But should the RDF spec change some day wrt the semantics of bNodes, then bNode semantics in OWL-Full would automatically change, too, without the need of having a OWL-WG to do this. Also, if an OWL-1.1-Full user or implementor does not want to cope with existential semantics, she doesn't have to investigate every point in the Full spec in order to see whether she will break something more then just not getting the typical entailments resulting from simple RDF semantics. [FIXME!]

I haven't read the OWL-1.0-Full spec for this property yet, but I do not remember to have ever seen explicitly stated dependencies of semantic conditions on existential bNodes in the AS&S. I will have a closer look at this, when I find the time (Personal-Action, if I do not get an official one :)).

A problem for this approach will be that there will probably be bNode related Full-entailment tests in the testcases spec. Something to think about, perhaps those tests should not be made normative? Again, I will check, what the current state is in OWL-1.0-Full.

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