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


You see to be forgetting the motivation for having profiles in the  
first place, which is that they can be "more simply and/or  
efficiently implemented". The motivation for OWL RL is to provide  
*scalable* reasoning for very large data sets; it achieves this via a  
rule based implementation that is sound for the OWL Full semantics  
and is even complete w.r.t. query answering -- the key inference in  
the kinds of application at which OWL RL is targeted -- for many  
ontologies. This makes OWL RL an extremely attractive proposition in  
many applications.

Of course there are no miracles, and this performance benefit doesn't  
come for free. The cost is that an OWL RL reasoner will not be  
complete for all ontologies and all entailments. Application  
developers will have to consider if this is likely to be a problem  
for them and if they are willing to pay the (potentially very large)  
performance cost attached to use of other reasoners -- such as an OWL  
DL or OWL Full reasoner.

As I understand it, your proposed "solution" to this problem is to  
retain OWL RL Full, and make its semantics be defined by the rule set  
(which, as a result of a very poor decision we made that the 2nd F2F,  
is implicitly the case in the existing spec). This means introducing  
a 3rd semantics for OWL: one that is highly counter-intuitive and  
highly incompatible with both the DL and Full semantics (we can't  
state equivalence for a large class of ontologies as is the case  
between the DL and Full semantics). Moreover, this would not change  
in any way the behaviour of rule-based reasoners -- the only  
"advantage" is that they would be able to declare themselves sound  
and complete w.r.t. this semantics.

In contrast, the proposed "unification" solution is to simply make  
the semantics of OWL RL be the OWL Full semantics. This avoids  
introducing a 3rd semantics, and allows us to relegate OWL R DL to  
simply defining a set of ontologies for which rule based  
implementations are sound and complete for query answering. This  
would not change in any way the behaviour of rule-based reasoners --  
the only "disadvantage" is that rule based reasoners would not be  
able to declare themselves sound and complete for all entailments and  
on all ontologies.


On 12 Aug 2008, at 01:08, Michael Schneider wrote:

> Hi Ian!
> Ian Horrocks wrote:
>> Michael,
>> It has never been claimed that OWL R DL and OWL R Full are completely
>> equivalent on the syntactic fragment.
> Ok. But still, we call it a "unification" of two languages. FWIW,  
> that's really not how I understand this term.
> Indeed, at the time when the issue was raised, I was under the  
> impression that the idea was that the two languages are  
> semantically equivalent on the syntactic fragment. That's why, in  
> one telco, I asked Boris whether this is really true, given that  
> the two languages look so different (to me). Because for me,  
> talking about "unification" only makes sense, when the two  
> languages to be unified are equivalent at least on the common part  
> of their respective syntax.
> And I also thought at that time that outside the common part, it  
> will be the rules which define the language. In this case, the new  
> language "OWL R" would have been a truly unified language, where  
> both languages, OWL R DL and OWL R Full, wouldn't have experienced  
> any disadvantages by the unification process. This situation would  
> have been perfectly ok for me, well, I would have regarded it to be  
> an improvement.
> But now, a reasoner, which was formerly sound and complete w.r.t.  
> the OWL R Full semantics (i.e. w.r.t. to what can be derived by  
> means of the ruleset), will suddenly experience heavy  
> disadvantages. For ontologies from the syntactic fragment, the  
> reasoner suddenly has to compete with the much stronger OWL 2 DL  
> semantics. And for RDF graphs outside the syntactic fragment, the  
> reasoner suddenly has to compete with the much stronger OWL 2 Full  
> semantics. No chance! The same reasoner can now only be said to be  
> a sound, but not complete OWL R reasoner. On the other hand, an old  
> sound and complete OWL R DL reasoner will still be OWL R sound and  
> complete, AFAIU. This wouldn't look like a very fair deal to me.
> In the original issue, it was stated that
>   "The main benefit would be that we would not need  
> owl:intendedProfile"
> For me, this is at best a very minor nice-to-have benefit. And now,  
> we are near to close this "intended profile" issue, anyway, by not  
> having such a signaling URI at all. So, the "unification" issue can  
> even be regarded to be kind of moot.
> But we are still talking about the unification, for which the price  
> to pay would be pretty high for the RDF side, which originally was  
> the only side that asked for such a rule-based language.
>> What is claimed (see Section
>> 4.4 of the existing Profiles doc) is that for ontologies within the
>> syntactically defined fragment the OWL RL rules will be complete
>> w.r.t. certain kinds of entailment -- in particular w.r.t. those
>> entailments that correspond to query answering.
> In this case (and I already suspected something like this, see my  
> mail [1] before the F2F) this semantic relationship is simply too  
> weak to be usable as a foundation for a unification. As long as the  
> two languages are strictly separated (the current status), it is  
> nice to know that there exists at least /some/ semantic  
> relationship (although I still do not see why it has to be stated  
> in the spec). But for a real "unification", which should try hard  
> to ensure that none of the unified languages experience heavy  
> disadvantages, this is not appropriate, IMO.
>> Your example is a good illustration of why it would be *a very bad
>> idea* to define a 3rd semantics for OWL based on the OWL RL rules.
>> According to this semantics, it would NOT be the case that
>> owl:intersectionOf (C D) is a subClassOf D. Any reasoner finding this
>> entailment would be unsound and non-conformant w.r.t. this semantics.
>> This would, IMHO, be highly counter-intuitive.
> Now this was actually my counter example, so one can easily take it  
> as an example for an "unintuitively" missing entailment. But  
> equally well, from a rules perspective, one could also claim that  
> producing this entailment is counter intuitive. Actually, there are  
> enough examples for derivations by applying the rules, where there  
> is no respective entailment by OWL R DL, simply because it would  
> fall outside OWL R DL's syntax (at least, an OWL R DL reasoner  
> wouldn't be required to produce it there).
> Consider that very asymmetric syntax of OWL R DL (e.g., unions, and  
> existential and universal restrictions may only occur on one side  
> of subclass axioms, respectively). Isn't this alone already  
> "counter-intuitive", at least to people who do not understand the  
> theoretic background behind this language? In comparison to this,  
> the OWL R Full ruleset looks pretty coherent to me.
> Well, so we have claims about "counter-intuitive" reasoning results  
> on both sides. I would call that a draw! :)
>> Moreover, I expect
>> that many rule based implementations *will* find this entailment, or
>> at least answer "yes" to a query asking if it holds, because they
>> will use the standard rule-based technique for answering such a
>> subsumption query, i.e., assert a fresh individual to be an instance
>> of owl:intersectionOf (C D) and check if it is entailed that this
>> individual is an instance of D. This entailment *will* follow from
>> the rules.
>> Regarding the guarantees that are made, guaranteeing soundness and
>> completeness w.r.t. OWL RL Full semantics for ontologies in the
>> syntactic fragment and certain kinds of entailment seems to be the
>> best that we can do -- as you yourself have clearly illustrated.
> I would probably agree that this is the best we can do, if we were  
> demanded to perform the unification. But at the moment, we still  
> have to decide *whether* to unify at all. And after this  
> discussion, I think that it will be the best to simply close this  
> issue without any further action. This will also be the solution  
> with the least effort, considering that the internal reviews are  
> going to start soon.
>> I don't pretend to fully understand your statements about reasoners
>> based on subsets of the OWL RL rules, but they seem to be based on a
>> misunderstanding about what is being stated w.r.t. soundness and
>> completeness, and on some strange assumptions about how conformance
>> will be defined.
> The point is that it would really not be a great marketing  
> statement to say: "We have a sound OWL RL reasoner!". Building  
> sound-only reasoners is a trivial task (just take a "zero- 
> reasoner", which does not produce any inferences). And a reasoner,  
> which just implements the "official OWL R ruleset", wouldn't be  
> more than just an OWL R sound-only reasoner. While there would be  
> other reasoners around, which would really be sound and complete  
> w.r.t. the OWL R semantics. Would anyone buy the former reasoner  
> under these circumstances?
>> Regards,
>> Ian
> Best regards,
> Michael
> [1] < 
> 0367.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:
> Web  :
> 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 Tuesday, 12 August 2008 11:20:19 UTC