Re: Asserting subclasses of open ranges or domains

On Sep 8, 2009, at 10:12 AM, Bernard Vatant wrote:

> Hi Heiko
>
> Indeed, but this is exactly what is *not* wanted. We don't want to  
> *restrict* the range to those classes, only explicitly *allow* them.
> I do not quite get this point. If the range is not restricted, then  
> everything is implicitly allowed.
>
> Yes, it is in the current state of affairs. But as DC comments puts  
> it, the range is not restricted not because it is by nature  
> completely open, but because DC doesn't know what the restrictions  
> should be. So it's open default of better definition.
>
> What extra information is provided by explicitly allowing something?  
> Consider the two statements "Books can be about every topic" and  
> "Books can be about every topic and about dinosaurs" - aren't they  
> equivalent? Or did I miss something here?
>
> Yes, in the current stae of affairs, you're right, from a purely  
> logical viewpoint, nothing is added by specifying subclasses of the  
> range.  Let me take another example. Suppose you want to set an  
> ontology of home appliances. You define a class "Home" and a  
> property "hasAppliance".
>
> Of course you can define the range of this property as  
> "DomesticAppliance", but you don't give any necessary condition for  
> this class, because you don't want to preclude any kind of stuff one  
> can invent there. So this class has a quasi tautological definition,  
> it is the range of "hasAppliance". No more, no less.
>
> Now I can, not restrict this class, but say which kind of stuff is  
> currently considered as domestic appliance. So I will define  
> "WashingMachine", "HomeCinema", "BathTub" and whatever you like as  
> subclasses of "DomesticAppliance". In short I have a list of  
> sufficient conditions for this class, but no proper necessary  
> conditions, beyond the tautological fact of being the value of  
> "hasAppliance". Sure enough, if the ontology had been set a century  
> ago, new subclasses would have emerge since, and an ontology set  
> today does not want to preclude whatever the 21st century will bring  
> about in this domain.
>
> We could as well define a tautological class dcterms:Topic as the  
> range of dcterms:Subject, and assert only subclasses.
>
> Is that clearer?

What is not clear is why you want to do this. Even in the case of the  
domestic appliances, if you do not put any necessary conditions on  
this class, you have effectively said nothing. It is tricky to appeal  
to intuition in cases like this, because of course we all know that  
there are things that are not domestic appliances, and we tend to use  
this knowledge without being told that we have to. But our ontologies  
only know stuff like this if we somehow tell them it explicitly.

The DC comment you appeal to is an indicator that, as you say, the  
"actual" range is not universal, but its extent has not yet been  
formalized. But asserting subsets of it - adding sufficient conditions  
- does not help any towards formalizing the actual range, even  
partially. The only thing that will do that is to somehow restrict the  
range with necessary conditions.

Pat Hayes

>
> Bernard
>
> -- 
> Bernard Vatant
> Senior Consultant
> Vocabulary & Data Engineering
> Tel:       +33 (0) 971 488 459
> Mail:     bernard.vatant@mondeca.com
> ----------------------------------------------------
> Mondeca
> 3, cité Nollez 75018 Paris France
> Web:    http://www.mondeca.com
> Blog:    http://mondeca.wordpress.com
> ----------------------------------------------------

------------------------------------------------------------
IHMC                                     (850)434 8903 or (650)494 3973
40 South Alcaniz St.           (850)202 4416   office
Pensacola                            (850)202 4440   fax
FL 32502                              (850)291 0667   mobile
phayesAT-SIGNihmc.us       http://www.ihmc.us/users/phayes