W3C home > Mailing lists > Public > public-owl-dev@w3.org > April to June 2007

Re: RDFS compatibility information in OWL-DL documents

From: Bijan Parsia <bparsia@cs.man.ac.uk>
Date: Tue, 17 Apr 2007 18:49:18 +0100
Message-Id: <40CF49E1-25E0-42DF-9273-BB6E3FFA3CE7@cs.man.ac.uk>
Cc: Alan Ruttenberg <alanruttenberg@gmail.com>, public-owl-dev@w3.org
To: Jeremy Carroll <jjc@hpl.hp.com>

On Apr 17, 2007, at 5:30 PM, Jeremy Carroll wrote:

> My view is that annotations are a shared feature between RDF and  
> OWL, and doing annotations in OWL the same way as in RDF is likely  
> to be a win.

That seems shared :)

> Hence, permitting RDFS reasoning, in full, on annotations seems  
> appropriate, for maximal interoperability with RDFS based tools,  
> for example, for display of data etc.

I consider the issue of what *sort* of reasoning/query/manipulation  
to support on annotations to be a bit, but not entirely, distinct  
from the question of how to, roughly, serialize annotations. Some  
users may want OWL reasoning, in full, on annotations. Others may  
prefer some logic programming like thing.

It's worth distinguishing the means from the effect. Consider the use  
of subproperty axioms with rdfs:label on the right hand side vs.  
using Fresnel selectors. These can have the same effect (i.e., the  
value of a certain property gets rendered in a UI). The typical  
fresnel lens will be described in a separate document, whereas, at  
least in RDFS land, the subproperty axiom will be in the schema (and  
ultimately in the dataset).

For the specific purpose of UI rendering, I prefer the general  
Fresnel approach, though I've had my issues with Fresnel. In Swoop,  
we made use of label information (and multilingual labels)  
fruitfully, but I've never felt the need to use rdfs:label as a  
selector of other properties as labels. So my personal best practice  
recommendations roughly say, use rdfs:label directly for convenience  
(e.g., when your URIs are not human readable), but use a more  
sophisticated mechanism when you want to use domain properties as a  
label (e.g., ex:surname). YMMV.

> An issue is the annotations on axioms, using reification in the  
> current OWL 1.1 specs. I appreciate that this is semantically  
> correct - the annotation is about the axiom, and not about some  
> participant in the axiom; but I wonder whether the OWL 1.0 approach  
> of annotating a class or a property or an individual, may have had  
> the merit of greater interoperability - in an area where  
> functionalities like pretty-printing are more pertinent than reasoning

OWL 1.1 has both axiom and entity annotations. That is, axiom  
annotations are an extension of OWL 1.0 annotations. See:
	<http://webont.org/owl/1.1/owl_specification.html#4.4>

Entity annotations are insufficient for many purposes (provenance,  
axiom deprecation or other flagging, timestamping, etc. etc.)

Cheers,
Bijan.
Received on Tuesday, 17 April 2007 17:49:35 GMT

This archive was generated by hypermail 2.3.1 : Wednesday, 27 March 2013 09:32:54 GMT