Re: OWL1.1 APis

On Dec 20, 2006, at 4:25 AM, Enrico Franconi wrote:

> (just to add some irrelevant noise to your mailboxes...)

Oh joy! ;)

> On 15 Dec 2006, at 09:49, Bijan Parsia wrote:
>
>> On Dec 14, 2006, at 8:32 PM, Gerd Wagner wrote:
>>
>>>> Remember that consistency and entailment are closely related,
>>>> e.g., P  is entailed by O if O & ~P is inconsistent.
>>>
>>> Only for classical logic, but not for many non-classical
>>> logics that may be more relevant to KR than classical logic.
>>
>> Well this is so for RDF through OWL, which is, after all, what we  
>> are *very specifically* talking about.
>
> With care...

Well, in my defense I do have only the  "if" running in the opposite  
way from the deduction theorem.

> First of all, in RDF there is no negation, so the above can not  
> even be stated.

Ill formed XMLLiterals and well formed XML literals are contraries,  
so there's an abbreviated form of negation (wildly abbreviated).  
However, it is true that in RDF you can't bring them together in a  
contradiction. You can in RDFS and you can express (boring)  
contradictions in RDF with suitable datatypes (e.g., positive and  
negative integers).

All I really meant is that if you are considering entailment, you  
need to pay attention to the contradictions your formalism allows.  
RDF and RDFS are not paraconsistent.

The *point* I meant to convey to Dave is that contradictions add  
entailments in the RDF-OWL family of languages. Thus, an RDF  
entailment engine and an RDFS entailment engine acting on the very  
same graph will get different entailments (in general) and that is  
easily seen by the fact that a graph interpreted under RDFS semantics  
can, in fact, be contradictory (i.e., have no models) whereas this is  
not the case for RDF (without extra datatypes). It was just meant to  
be an "obvious" case that helped highlight the fact that the  
"nesting" behavior of current semantic web languages did not require  
that a reasoner for a more expressive (in some respect) logic *only*  
produce the exact same entailments for a graph that a reasoner for a  
less expressive logic would do for that graph.

When I get a moment, I'll map out the relationships between the  
various logics in RDF-OWL wrt entailment with examples to make it all  
as clear as I can make it so that people can understand how OWL 1.1  
fits in. Which is, after all,  the *point* of this thread.

(I agree that there are tricky bits.)

> Also in OWL-DL there is no negation in statements.

I presume you meant in facts/ABox statements?

Well, what is
	a:~C
?

I.e., in OWL abstract syntax:

	Individual( a type(complementOf( C))


Sure looks like negation in a statement to me!!!!

(Of course, OWL-DL facts are not closed under negation, i.e., you  
cannot just negate an arbitrary fact.)

> In DLs (like in modal logics - which are non-classical logics) the  
> deduction theorem is not straightforwardly true.

Good for me I didn't state the deduction theorem per se! :)

> As a matter of fact, the reduction from entailment to  
> satisfiability is done by using the internalisation process (e.g.,  
> by having or mimicking the universal role).

I guess my new new years resolution should be never to make a (even  
true!) unguarded/qualified statement :) (er...*EXCEPT* where, by a  
sort of "closed world" (closed mind? :)) assumption, I know everyone  
*really will* get the context and respect it accordingly :)).

Cheers,
Bijan.

Received on Wednesday, 20 December 2006 09:54:21 UTC