Re: ISSUE-131 (OWL R Unification): Fundamental understanding problems

Michael,

Most of your questions relate to conformance issues, something we  
didn't make any firm decisions on yet. The default assumption has  
been that conformance will be defined similarly as in OWL, i.e.,  
confirmant reasoners guarantee "correct" behaviour only for  
ontologies matching the relevant syntax specification, but this may  
need some refinement. I have commented in a little more detail below.

On 8 Aug 2008, at 01:23, Michael Schneider wrote:

> Dear all,
>
> in spite of a lot of discussion, which has happened on the unification
> issue, both in the mailing list and at meetings, and additional  
> discussions
> which I had privately with WG members, I stronger than ever have  
> the feeling
> that I do not understand what the unification will actually look  
> like in the
> end. So I would like to hear from other WG members what they  
> believe will be
> the result.
>
> Below, I have assembled a list of assumptions, which represent my  
> current
> understanding. These points can, in the simplest case, be answered  
> with
> "yes" or "no". But in the case of "no", I would appreciate to hear  
> from you
> what your understanding is instead.
>
> Please keep in mind that I am really only interested in  
> understanding how
> exactly OWL R will be specified in the end. This means,  
> essentially, that I
> want to know what the syntax and what the precise semantics of OWL  
> R will
> be, and also, what the exact role of the OWL R ruleset will be for  
> OWL R.
>
> (On the other hand, I am *not* interested in, for example, the  
> question when
> may or should an OWL R reasoner signal a warning to a user, since  
> such an
> implementation-specific behavior is completely outside the formal
> specification of a language.)
>
>
> Here is the list of my current assumptions:
>
> (A) The "syntactic fragment" of the unified OWL R language will be  
> defined
> by today's syntax of OWL R DL, as specified in sec 4.2 of the Profiles
> document. An ontology will be called a "valid OWL R ontology", if  
> it matches
> these syntactic restrictions.

This is a conformance issue --- it isn't yet determined how a valid  
OWL R ontology will be defined.

>
> (B) Nothing specific is said about ontologies which do not fall in  
> this
> syntactic fragment, so an OWL R compliant reasoner is free to  
> either deny or
> accept such an ontology as syntactically valid input.

Conformance again. My assumption is that an OWL RL reasoner will  
accept all RDF graphs, and that it may issue a warning for graphs  
that are outside the syntactic fragment.

>
> (C) For ontologies matching the syntactic fragment, the OWL R  
> semantics of
> such an ontology will be specified in parallel by the OWL DL  
> semantics (as
> it is nowadays true for OWL R DL), and also w.r.t. the ruleset of  
> today's
> OWL R Full.

No. The semantics of OWL-RL ontologies is the same as for every other  
OWL ontology, i.e., it is either the MT or the RDF semantics. The OWL  
RL rules do not define a semanitcs (at least we won't use them for  
that purpose) -- if they did, it would be equivalent to neither the  
MT nor the RDF semantics -- it would be a 3rd semantics for OWL (and,  
as you illustrated in your other email, a highly counter-intuitive  
semantics). However, for ontologies that are inside the given  
syntactic fragment, implementations based on the rules can give  
precise guarantees w.r.t. the MT/RDF semantics (which are themselves  
equivalent for this fragment).

>
> (D) For (C) to make sense, the DL semantics and the rule based  
> semantics
> have to be exactly equivalent for ontologies matching the syntactic
> fragment. It is believed that this relationship holds. For ontologies
> outside the syntactic fragments, this equivalence is *not* required  
> to hold.

As I mentioned, the OWL RL rules will not be used to define a  
semantics. It is believed that, for ontologies that are inside the  
given syntactic fragment, implementations based on the rules can give  
precise guarantees w.r.t. the MT/RDF semantics. Note that this is NOT  
the same as saying that such implementations will be "semantically  
equivalent" to OWL 2 Full on this fragment -- the guarantees are only  
w.r.t. certain kinds of entailment (as already defined in Section 4.4  
of the Profiles document). For ontologies outside this fragment such  
guarantees cannot be provided in general.

>
> (E) For ontologies outside the syntactic fragment, the only semantic
> restriction on reasoners is that they must not produce inferences  
> which go
> beyond OWL Full (without "R"!) entailment. So they may produce  
> whatever
> inferences they like, as long as they keep being in the scope of  
> OWL Full.
> In particular, they MAY produce all or only some of the inferences  
> which can
> be derived from the OWL R ruleset for such ontologies, but this  
> will in no
> way be enforced by the specification or OWL R.

Conformance again. My assumption is that for ontologies outside the  
given syntactic fragment OWL RL reasoners will be sound w.r.t. the  
RDF semantics (i.e., won't produce any more entailments than a  
conformant OWL Full reasoner) and not less complete than  
implementations based on the OWL RL rules (i.e., will produce at  
least those entailments that would be derived using the OWL RL rules).

Regards,
Ian



>
>
> Regards,
> 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 Monday, 11 August 2008 10:29:09 UTC