Re: Question about domains and rages

Hi Andre Luiz,

Please find some clarifications below...

On 05/06/2013 05:37 PM, André Luiz Tietböhl Ramos wrote:
> Hello,
>
> Being a beginner in Protégé, I'd like to understand fully the concept 
> behind domains and ranges which wasn't clear to me after studying its 
> tutorial.  Suppose, as in my case, that I have an object property 
> named hasComponentNeed.  It describes the type of component that is 
> required by other components in the ontology. Likewise, there is its 
> inverse & symmetric property called isComponentNeedOf.

I don't think isComponentNeedOf should be a symmetric property (which 
BTW would make its inverse property, hasComponentNeed, also symmetric).

This would mean that if c1 isComponentNeedOf c2, then also c2 
isComponentNeedOf c1, which conform to my intuitive interpretation of 
the meaning of these properties would not be the case. I would even say, 
it would NEVER be the case, which would make these properties 
asymmetric, rather than symmetric.

In the hope that it would make the meaning of these properties more 
clear, I would suggest to rename them as follows:
hasComponentNeed  -->  needsComponent
isComponentNeedOf  -->  isNeededBy  (or isNeededComponentOf)
I hope that I am not misinterpreting their meaning.

In this case it is obvious that if component (or system) c1 needs a 
component c2, then c2 is suppose to be a *proper part of* c1, which is 
an asymmetric relation.

>
> I'm adding the components' classes as the object property domain of 
> hasComponentNeed. 

I don't know how you should model the domain of hasComponentNeed, but I 
would definitely add Component as the *range* of the 
hasComponentNeed/needsComponent property.

> The issue I'm concerned about is that the object property domains 
> state it is the intersection, as it does for the ranges.  Should I 
> understand the domains and ranges intersection requirement in the set 
> theory context?  My humble understanding about this is that the 
> domains work more as a "container" for classes.

Yes. [1] states that if you have multiple domain/range axioms for a 
given property in an ontology, then the domain/range must be interpreted 
as the conjunction of those classes. Indeed, the intersection is meant 
in the set theoretical interpretation. For example, if you state that
     the domain of hasDriverLicenseNumber is Person,
and you also state that
     the domain of hasDriverLicenseNumber CarDriver
it means that the real domain of hasDriverLicenseNumber is (Person 
intersection CarDriver). This means that people who cannot drive a car 
(i.e. are not car drivers), cannot have a driver license number. Also, a 
monkey from a circus that was trained to drive cars, and therefore one 
could assert that it is an instance of a CarDriver class, can not have a 
driver license number, because it is not member of the Person class.

It is very important to understand that stating the domain D of a 
property p does not "attach" that property to the class D, in the sense 
that all individuals of class D would have a property p, but rather it 
would say that if there is a statement (ind p val), then a reasoner can 
conclude that individual ind is of type D.
I suppose that this is different from what you thought about a domain 
behaving as a "container" for classes.

As [1] nicely describes, if one would like to have multiple classes in 
the domain/range of a property, one should use an expression that 
describes the union of those classes, instead of writing separate axioms.

>
> When it infers the ontology, the ranges are reasoned right.  Is this 
> correct?

I am not sure what you mean by this, but I suppose that if it is 
reasoned right it must be correct :)

>
> I appreciate any help.
>
> Thanks,
>
> -- 
> Andre Luiz Tietbohl Ramos
> http://www.feng.pucrs.br/~andreltr <http://www.feng.pucrs.br/%7Eandreltr>
>
> //
>
>

Regards,
Csongor

[1] http://www.w3.org/TR/owl-ref/#domain-def

_______________________________________________
protege-discussion mailing list
[email protected]
https://mailman.stanford.edu/mailman/listinfo/protege-discussion

Instructions for unsubscribing: http://protege.stanford.edu/doc/faq.html#01a.03