SKOS vs OWL (was: Threads)

[Norman Gray]

[This is a response to a number of other messages in this thread]

On 2007 Sep 28, at 08:20, Andrea Preite Martinez wrote:

> On the other hand, I think that the discussion of vocabularies AND  
> ontologies, of SKOS or OWL, is perfectly irrelevant to the present  
> user requirements.

The problem with 'SKOS or OWL' is that a vocabulary isn't an  
ontology.  Adding "<> a owl:Ontology", and changing skos:Concept to  
owl:Class doesn't make it one.

The super-/sub-class relations in an ontology are required to be  
homogeneous, in that all instances of a subclass are necessarily also  
instances of the superclass, and that simply doesn't work for the  
'kinematics' NT 'acceleration' example I mentioned earlier.  Avoiding  
expressing those, and just adding a lot of SKOS terms produces a  
strange mutant: a SKOS vocabulary which has a few OWL terms in it,  
adding some syntactical baggage without (as far as I can see) doing a  
lot of work.



[The next three paras are formal minutiae]

There's also a technical difference, which is that the things which  
are of type skos:Concept are specifically 'concepts', not other  
things.  Thus an OWL class myns1:Star is (probably) the class of  
things which are stars; a skos:Concept myns2:Star is an _instance_  
which names 'the concept of a star'.  The skos:Concept  
myns2:VariableStar is not a subclass of myns2:Star, because neither  
of them is a class.  This sounds nit-picky, but is the suggestion  
that we'd get ourselves in trouble one way or another -- we'd end up  
saying something nonsensical and deduce rubbish -- if we conflate the  
two.

There's an example at the end of the SKOS spec which distinguishes  
the dc:creator of the concept of "Henry VIII" (which might be me,  
today), from the dc:creator of "Henry VIII" (the person/instance,  
which was done by "Henry VII" and "Elizabeth of York"), and neither  
of those things is the class of things which are Henry VIII (now get  
an aspirin).

OK, so talk about the owl:Class myns1:Star and develop that in  
parallel with the skos:Concept myns2:Star, perhaps linking them with  
something like myns1:hasAssociatedConcept.  But then you're  
developing a vocabulary and an ontology both at once, and it's not  
clear (to me) what work the ontology half is doing.



Short version: I don't see the Requirement for an ontology.

All the best

Norman


-- 
------------------------------------------------------------
Norman Gray  :  http://nxg.me.uk
eurovotech.org  :  University of Leicester, UK