RE: XG W3pm Scope

 wrt your last point ("hasValue calculation"):
I have along background in (mostly closed world) modelling stuff like ISO STEP EXPRESS, UML, etc.
This means that sometimes when modelling in RDF(s)/OWL I interprete things wrongly or apply things wrongly....(just by habit). I think the assumption that "a hasValue should be calculated to result in a valid individual" instead of "a check whether an individual HAS a certain value to classify" is typically such an example ...

Cheers, Michel
 
 

-----Original Message-----
From: Kendall Grant Clark [mailto:kendall@clarkparsia.com] 
Sent: 04 June 2008 14:19
To: Bohms, H.M. (Michel)
Cc: Henson Graves; public-xg-w3pm
Subject: RE: XG W3pm Scope

Excerpts from Bohms, H.M. \\\\(Michel\\\\)'s message of Wed Jun 04 03:29:34 -0400 2008:
> 
> 
> Dear Kendall,
> 
> Just some extra info on our BIM work, esp. the PMO upper ontology from SWOP:
> 
> In SWOP we use our own devised Rule Model explcitly modelling 
> assertions ("the stair width should be smaller than the toilet width 
> in that groundfloor of this house" etc. and derivations (kind of 
> scripts) like "the height of a façade is always 2/3 of the length of 
> that façade"; the first are checked the latter are executed (they can 
> all be mixed given quite some power and at the same time 
> end-user/modellers headache:)

OWL2 n-ary datatype predicates, as I presently understand them, allow the 2nd but not the 1st of yr examples. They have to apply to properties of a single individual -- an instance of, say, StandardFacade, -- rather than to different individual (instances of Toilet and Stair), though I suppose there might be a perverse modeling trick to get around that (well, maybe not -- have to think about it).

> Surely, we looked at existing stuff like SWRL but although we saw that 
> these things were way more formal/better defined they had too limited 
> power; hence our own more business related rule model.

Yes, a perfectly reasonable choice IMO.

> Clearly, if there are ways to NOT do it ourselves but move this 
> functionality to standard reasoners this would be our 
> preference...we'd like to saty as close to existing standard as possible.

I suspect OWL2 may give some means to move *some* more stuff into standard infrastructure, though I'd be surprised if you could get all the way there. I don't know if this is game where incremental wins are still wins or not. YMMV.

> If we not consider rule language/reasoners for that....there were also 
> issues for PMO wrt owl-level reasoners like eg for the automaitic 
> execution of hasValue constraints. For example we can define variant 
> classes for products that have a fixed (class-level) value; eg a 
> catalogue item having a price in euro's. In our system this is still a 
> class that can have many individuals all having the same price value. 
> Clearly we can specify in owl that the price is always x euro's but you need some reasoner to actually MAKE it that value.

I'm not sure, but I think this gets toward integrity constraint semantics, where a class description acts like a template or pattern that thing must satisfy, rather than a way for the reasoner to infer new information.

Cheers,
Kendall
This e-mail and its contents are subject to the DISCLAIMER at http://www.tno.nl/disclaimer/email.html

Received on Wednesday, 4 June 2008 13:06:09 UTC