Re: OWL-S process preconditions

On Jun 22, 2004, at 11:32 AM, Gerhard Wickler wrote:

> Bijan Parsia wrote:
>
>> On Jun 22, 2004, at 8:41 AM, Gerhard Wickler wrote:
>>
>>>
>>> Good point. I agree that having RDF/XML as the content language 
>>> (using reification) looks like a good idea, especially because it 
>>> answers my semantics question,
>>
>>
>> No, that it doesn't but it's tempting to thing it does is precisely 
>> why it's a bad idea, IMHO.
>
>
> It would indeed be a bad idea if RDF/XML did not have a formal 
> semantics,

RDF does have a formal semantics.

>  but it does, doesn't it? To be honest, I haven't looked at the 
> semantics document in great detail, but it does seem to define a 
> model-theortic semantics the way I'd expect it to.

But it's completely insufficent to express anything like preconditions 
*in* RDF. So, adding more triples doesn't get you any more juice. 
Reificiation is particularly underspecified.

So any rules/predconditions/planning language is going to have to build 
its own semantices outside of RDF. The precise relation of a new 
language's semantics (and syntax) is a matter of considerable debate. I 
suggest checking out Pat Hayes's and Peter Patel-Schneider's 
presentations at the Boston Tech plenary...uhm...in 2003?

Ah, here's a link:
	http://www.w3.org/2001/sw/meetings/tech-200303/

I see links there to the LBase document, too.

"Three these of represenation  in the Semantic Web" also discuss these 
issues:
	http://www2003.org/cdrom/papers/refereed/p050/p50-horrocks.html

Some may find it contentious, of course.

Think of it this way. Propositional logic has a model theory. Yay. We 
like that. We can encode quantified formulas in propositional logic 
(e.g., by reification, i.e., propositions about the structure of 
various formulas). But we'd all hate it fiercely. Oops, wait, I mean, 
we wouldn't think that that encoding alone grounded the meaning of our 
pquantified formulas in any interesting way.

The LBase approach is a little different because you translate to a 
*richer* language and then add bridge axioms to that language's 
constructs that have the appropriate meaning. For many purpose, this is 
still painful. It also can be practically difficult, depending on what 
sorts of constraints you want to impose on your tools.

> Anyway, we appear to agree that it is not a good idea to use RDF/XML 
> with reification as a preconditions/effects language.

Well, though I pushed for using literals, I don't care *so* much. I was 
worried that if we used the standard SWRL encoding, there'd be 
ambiguity between SWRLish use and OWL-Sish use. OTOH, revising all the 
existing SWRL tools is trivial :)

I think it's also a bit more extensible, in that it's more natural to 
just *use* the canonical syntax of alternative precondition languages 
(e.g., kif). But you still have to do *some* work, e.g., to harmonize 
variable bindings.

Cheers,
Bijan Parsia.

Received on Tuesday, 22 June 2004 11:58:19 UTC