Re: OWLED 2007 is nigh

On 30 May 2007, at 14:29, Dan Brickley wrote:
[snip]
> I'd be very interested to hear what owners of publically-used OWL  
> namespaces hope for from the W3C rules / RIF work, eg. whether  
> publishing some kind of RIF data via their namespace documents is  
> something the OWL community care about.

I'm not sure what you mean, esp. as I don't know what an "OWL  
namespace" is.

> And in general, pet-peeve-wise ... to get a sense of issues around  
> integration across W3C standards (SPARQL/RIF/OWL/SKOS/GRDDL/RDFa/ 
> etc) as they're seen from an OWL perspective.

Hmmm. Do you mean do people want to publish a RIF *variant* of their  
ontologies?

I'm very unclear how RIF will shake out and whether "Exchange" means  
you should publish in it.

At OWLED, I hope that we will agree to start putting together a spec  
for DL Safe rules as an extension to OWL. For the fragments of OWL  
which have reasonable readings and reasoning over as horn clauses  
(e.g., DLP and hornSHIQ) it would make sense to define RIF dialects  
for them (or markers in subsets of RIF dialects). Bit premature on  
that front as RIF is still getting sorted on basics.

> I'd also like to hear whether there are any emerging conventions  
> for ontologies that have an OWL-Full form to also publish an OWL-DL  
> version (eg. content negotiation? linked from namespace, etc). I  
> know for eg that DL-ized versions of Dublin Core and FOAF are  
> floating around, there are probably others too. What can be done to  
> improve that situation, beyond migrating everything to pure DL?

My personal strategy, which is shared by several people and certain  
can be seen in the design of OWL 1.1, is to try to eliminate the  
distinction between OWL Full and OWL DL as much as possible by  
incorporating as much of OWL Full features as possible in, or  
providing effective alternatives when possible. Punning is one  
example, and while I'm not satsified by it over all, our experience  
with Pellet shows that an awful lot of OWL Full documents can be  
usefully and perhaps correctly processed that way. Similarly for  
subpropertying rdfs:label. In general, I think it's good to separate  
out the annotation requirements and the domain modeling requirements  
quite a bit, e.g.:
	<http://www.w3.org/mid/CE5ED511-BF3F-4C01- 
BC75-9D6AA4F551A3@cs.man.ac.uk>

> Sounds like a good event, sorry I can't be there...

It will be. Me too.

(It's probably better to discuss the technical details of these on  
public-owl-dev.)

Cheers,
Bijan.

Received on Wednesday, 30 May 2007 17:07:03 UTC