[protege-owl] "Abstract" class in OWL

[Thomas Russ]

On Sep 30, 2007, at 4:31 PM, Fernando Mertins wrote:

> Hi...
>
> Is there an option to define an "abstract" class, in OWL? i.e., a  
> class that
> must not have instances... The propose of the class is only to be  
> superclass
> from others... i.e.:

Not directly.  It doesn't really make sense in OWL, because the  
notion of what class something belongs to is a lot more fluid in  
OWL.  This can change with additional assertions or inference, so  
there isn't the same degree of concern as with object-oriented  
programming.

>
> Model
>   Model A
>   Model B
>   Model C
>
> A, B and C classes are "models", but something cannot be only  
> "Model", must
> be a model of subtype A, B or C...

You could make "Model" equivalent to a union of "Model A", "Model B",  
and "Model C".

That would mean that anything that is a Model must also be one of  
Model A, B or C as well.  But since OWL is quite happy with  
incomplete (open world) knowledge, you could have some particular  
model about which you didn't know which of those subclasses it  
belongs to.  I also note that if you don't want to have something  
possibly being both a Model A and a Model B, etc. you will have to  
tell OWL that A, B and C are disjoint classes.  By default they are not.

One key thing to understand when coming to a description logic from a  
programming background is that DL is about inference and not about  
primarily about type-checking.  The inference and the different  
assumptions underlying the reasoning system can cause some surprises  
until you adjust your way of thinking.