Re: Additions to scope of OWL1.1

On Nov 23, 2006, at 8:55 PM, Hans Teijgeler wrote:

> Bijan,
>
> Is the scope definition of OWL1.1 closed, or are proposals for  
> additions still possible?

It is not exactly closed, but it's not wide open either. If you see  
my weblog post from last year about it:
	http://clarkparsia.com/weblog/2005/12/20/owl-11/

I hope that makes clearer the criteria we used: theorists and  
implementors had a clear idea that the proposed feature was doable  
and there was strong consensus from the users that *exactly* that  
variant of the feature was wanted and would be used. Moreover, the  
implementors had to actually agree to implement it :) (Some were  
easy, like qualified number restrictions. Racer and FaCT++ already  
had support and it wasn't too bad at all to add it to Pellet.)

These are tough bars! But we wanted to *change* the de facto standard  
(and eventually the de jure standard!) and this requires getting  
broad agreement. Many people are skeptical of ANY changes, and more  
are skeptical about adding expressivity. So, the intent is careful  
stepwise refinement.

Obviously, either "too fast" radical changes or *no* changes can be  
destabilizing or harmful to continued growth and success of OWL. So  
we are trying to thread our way toward handling unmet needs without  
causing marketing problems.

Standardization is also "expensive" both in money but also in will,  
energy, etc.

>  In the world of ISO 15926 we are struggling with the problem that  
> in RDF we cannot model something like:
>
>     :Joe isFatherOf :Mary
>     :Mary isInvolvedWith :Pete
>     :Joe approves [:Mary isInvolvedWith :Pete]
>
> Often it is necessary to say something about a relationship between  
> two individuals (relations involving relations, like above  
> example), with the requirement to constrain that in an OWL schema.

*My* thought is that this is too much for OWL 1.1. It is *certainly*  
too much without a very specific worked out semantics (even informal).

However, there is a way to put annotations on relations:	
       http://owl1_1.cs.manchester.ac.uk/owl_specification.html#6.4

This would allow you at least to *syntactically* express what you  
want. That could serve as a basis for figuring out a semantics that  
works for you.

I encourage you to write up your requirements with lots of examples  
and some idea of what you want it all to mean and submit it to  
OWLED2007 (CFP coming soon!). Even if it's not quite ready for de  
jure standardization, you may be able to interest people in  
implementing it in a consistent way. I have some work started in this  
direction, and I'm interested as to whether it'd help.

> ISO 15926 uses classes for (binary) relationships. These classes  
> have two properties, the ranges of which coincide with the domain  
> and range of an owl:ObjectProperty. For example:
>
>     :Mary isInvolvedWith :Pete
>
> maps to:
>
>     <Involvement rdf:ID="ME329800">
>         <involved1 rdf:resource="#Mary"/>
>         <involved2 rdf:resource="#Pete"/>
>     </Involvement>
>
> where any subsequent approval of this relationship can be done like:
>
>     <Approval rdf:ID="ME324199">
>         <approved rdf:resource="#ME329800"/>
>         <approver rdf:resource="#Joe"/>
>     </Approval>

Yes, reification is a standard trick. If this captures your meaning,  
then it may be that some judicious syntactic sugar is all you need.  
That would be at least reasonable for OWL 1.1.

> We could use an ObjectProperty AND a Relationship class, but then  
> we have no way to tell that a particular instance of the one is  
> semantically equivalent to a particular instance of the other. It  
> would require a.k.a. cross-over between equivalentClass and  
> equivalentProperty.
>
> This may be way out of line for the OWL1.1 exercise. In that case I  
> apologize for taking your time.

No need to apologize! I'm *always* interested in hearing about  
interesting needs. I'd be happy to help evolve a description of your  
needs or to help craft a proposal that might get some traction.

I certainly hope and expect that OWL 1.1 is not the *last* step in  
the evolution of OWL.

Cheers,
Bijan.

Received on Friday, 24 November 2006 07:55:43 UTC