W3C home > Mailing lists > Public > public-owl-wg@w3.org > April 2008

Re: General discussion for TC Wednesday 2008-04-21

From: Bijan Parsia <bparsia@cs.man.ac.uk>
Date: Tue, 22 Apr 2008 09:46:21 +0100
Message-Id: <6A99CE06-C5ED-4EBB-9AC6-34DD99FB089B@cs.man.ac.uk>
Cc: OWL Working Group WG <public-owl-wg@w3.org>
To: Alan Ruttenberg <alanruttenberg@gmail.com>

On 22 Apr 2008, at 09:02, Alan Ruttenberg wrote:
>
> On Apr 21, 2008, at 5:33 PM, Bijan Parsia wrote:
>> On Apr 21, 2008, at 10:13 PM, Alan Ruttenberg wrote:
>>
>>> A quick glance at the document cited reveals only an XML syntax,  
>>> and the RIF/RDF/OWL documentation[3] suggests that combinations  
>>> of RIF and RDF be made by way of using multiple documents.  Is  
>>> there a RDF syntax proposed, and if not are we comfortable with  
>>> having key axioms specified in a non-RDF syntax.
>>
>> I've put the proposal into the Wiki.
>
> Ok. Thanks for that, and sorry about the confusion.
>
> BTW, I was wondering whether the proposal could be extended to  
> include very easy keys for classes. This covers a use case I have,  
> similar to the other bioinformatics ones, except where the entities  
> are represented as classes.
>
> It would be sufficient to have class keys be global. Same keys  
> would imply equivalentClasses as opposed to sameIndividuals
>
> classKeyPropertyExpression := dataPropertyExpression |   
> objectPropertyExpression
> classKeyAxiom := 'ClassKey' '(' { annotation} keyPropertyExpression  
> { keyPropertyExpression }')'
>
> e.g.
>
> ClassKey(keggId)
>
> I think this could be accomplished as a preprocessing step to  
> compute the equivalent classes before any (harder) reasoning starts.

I don't think the extension is straightforward (indeed, getting the  
easy keys we have was surprisingly unstraightforward), primarily  
because it involves a bit of higher-orderness, which would require  
another extension to DL Safe rules (or the first order translation).

Perhaps the current proposal + hilog would suffice (i.e., the classes  
get their key properties in individual mode, then the keys force them  
to be equal, then hilog forces them to be equivalent). I don't know  
and will have to think about it.

Unfortunately, it's not obvious that a preprocessing step would work,  
esp. if the Hilog route is (more or less) correct. Disjunction and  
class expressions leap to mind as things that would need  
consideration. I do have an idea for "easy Hilog" but it's far from  
fleshed out.

So, worth thinking about, I think, but I'm cautiously pessimistic.

Cheers,
Bijan.
Received on Tuesday, 22 April 2008 08:44:26 GMT

This archive was generated by hypermail 2.2.0+W3C-0.50 : Tuesday, 22 April 2008 08:44:28 GMT