W3C home > Mailing lists > Public > www-webont-wg@w3.org > February 2003

RE: Annotations and non-mon example

From: Jeremy Carroll <jjc@hpl.hp.com>
Date: Sat, 1 Feb 2003 21:26:40 +0100
To: www-webont-wg@w3.org
Message-Id: <200302012126.40115.jjc@hpl.hp.com>

Jeremy (on OWL DL annotations)
>>Could I suggest it would be clearer with a new rdfs:Class
>>   owl:AnnotationProperty
>>and we require all annotation properties to be of this class.
>>This is not my preferred solution

>I could live with this 

> (in fact, as a tool builder it would be useful
>- because we could use subclasses of annotationProperty
Aside: that would put you in OWL Full, where owl:AnnotationProperty would be 
irrelevant, and fairly vacuous.

> well, I just can't for the moment

I hope I have been sufficiently clear.

This was suggested only as a clarification of what I think the current AS&S 
text is trying to say about OWL DL (and OWL Lite).

It certainly would not help with the underlying semantic failure that AS&S 
treatment of annotations is simply wrong.

My reading of what AS&S currently says about properties in OWL DL is that 
there are four sorts of property used in OWL (other than the built-ins):

- datatype properties 
  These have type owl:DatatypeProperty
- object properties which may be a transitive or have a syntactic super 
property (following subPropertyOf, samePropertyAs and inverseOf links) that 
is transitive.
  These have type owl:ObjectProperty
- object properties which may have cardinality constraints on them or their 
subproperties (understood syntactically as above)
  These too have type owl:ObjectProperty
- properties used as annotations
  These have no type whatsoever (they may not have type rdf:Property).

I think AS&S's "side condition" about transitive properties is an inadequate 
expression of the syntactic constraint that holds between 
owl:TransitiveProperty and cardinality restrictions; which precisely because 
of its subtly is best expressed in syntactic rules.

The owl:AnnotationProperty idea comes from seeing it as a mistake to have a 
distinct syntactic category that is defined in the concrete syntax by an 
absence of information rather than by the presence of information.

My preferred solution merges the syntactic category of annotation property 
with the other syntactic categories of properties - ending up with five 
different categories of property in the abstract syntax, but only two markers 
(owl:DatatypeProperty and owl:ObjectProperty) in the concrete syntax.


Received on Saturday, 1 February 2003 15:25:44 UTC

This archive was generated by hypermail 2.3.1 : Tuesday, 6 January 2015 21:56:51 UTC