Re: why are properties decoupled from classes

>> OWL allows you to define a cardinality on how a property applies to  a 
>> particular class. <http://www.w3.org/TR/owl-ref/#cardinality-def>
> 
> 
> I don't agree there (but maybe my interpretation is wrong, please  
> correct me in that case).
> 
> My reading of that description is:
> 
>     IF a person has a mother THEN it has exactly one
> 
> but you can have a person WITHOUT a mother and be consistent (not  
> because of the open-world assumption, but because the semantic of  
> owl:Cardinality does not imply that *for all*).

No it means that if a thing is a person then it has exactly one mother. 
Conversely it means that someone with zero or more than one mother is 
not a person.

To express that a thing could have either one or zero mothers you would 
use maxCardinality and minCardinality (with the values one and zero).

However being told that a person has a mother does not mean that you 
know anything about that persons mother, a graph which does not contain 
a triple with an _ex:biologicalMother predicate is still valid (for that 
matter so is one with two such triples pointing to different objects - 
it's only if you can somehow entail that those two objects are 
definitely not the same object that you can know that the graph is in 
error).

Received on Wednesday, 26 October 2005 14:50:28 UTC