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

Re: Annotations use case

From: Ian Horrocks <horrocks@cs.man.ac.uk>
Date: Wed, 12 Feb 2003 22:44:48 +0000
Message-ID: <15946.52832.800693.430554@merlin.horrocks.net>
To: Jim Hendler <hendler@cs.umd.edu>
Cc: Mike Dean <mdean@bbn.com>, webont <www-webont-wg@w3.org>

On February 12, Jim Hendler writes:
> 
> At 9:44 +0000 2/12/03, Ian Horrocks wrote:
> >Another point on annotation.
> >
> >I presume that it is obvious by now that we need to have annotations
> >in the RDF graph (XML comments just don't hack it as there is no
> >guarantee that they would be preserved when exchanging or editing
> >ontologies).
> 
> No, I think we agreed that we need to have annotations in the RDF 
> Graph.  I wouldn't agree to the parenthetical, but don't think it is 
> germane.

It is very germane. In any realistic application ontology authors are
going to want the facility to add comments that are not semantically
meaningful and which are guaranteed to persist as the ontology is
shared and edited using a variety of tools. Lack of tool
interoperability has already been identified as one of the main
weaknesses of DAML+OIL, and using XML comments would be fatal to hopes
of OWL tool interoperability.


> >  If all such comments are semantically meaningful, then
> >there is a serious issue with backwards compatibility. E.g., if I
> >correct a spelling mistake in an annotation, then is the resulting
> >ontology backwards compatible with the original?
> 
> No, there is no problem w/backwards compatibility -- if you change an 
> annotation you change the document -- how is the computer to tell a 
> significant from an insignificant change?

Because the author of the ontology tells it which are significant!
Those parts whose meaning is intended to be machine
understandable/processable are encoded in the logical language that we
have so carefully designed and whose meaning is well defined and
amenable to automated reasoning; those parts whose meaning is not
intended to be machine understandable/processable are placed in
comments/annotations that may be understandable to a human but not to
a machine.

>  If no annotations count, 
> you cannot address the use cases that started this thread (and Ian, 
> with due respect, you've just ignored them - how *would* you handle 
> the "embarass" case whcih I remind you, is an ability our 
> requirements necessitate)

I ignored them because they are irrelevant. You seem to believe that
inventing "meaningful" names for things gives them meaning. If this is
true then we don't need OWL or RDF - we can just use XML.

Take your embarrass example. The character string "Embarrass" has no
meaning whatsoever as far as a machine is concerned. So if a machine
finds that it can deduce "Embarrass", it wont feel either pregnant or
red in the face - it will simply be able to draw whatever conclusions
follow from "Embarrass". The language labelling just says how objects
should be displayed to a human, and need not be understandable by the
machine. If you want it to be understandable, then include it in a
restriction with a string valued datatype property, and not in a
comment or label.


> So fixing a spelling mistake may sound absurd, but consider if I have 
> a web page that points to
>   http://www.w3.com/.../owl
> and change it to
>   http://www.w3.org/.../owl

How is this relevant?


> I've made a real change to the site and, for example, a crawler 
> revisiting my site is likely to change a number of things based on 
> this.  We have a good solution to your problem -- when you fix the 
> comments, you add that your new version is backwards-compatible with 
> your old version so people know nothing contentive has changed.

I tell the machine that what I am changing is not contentful by
including it in a comment! I don't want to have to use another
(possibly meaningful?) annotation to indicate that changes to other
annotations are or are not contentful.


> Look, I realize there is a real issue here - we have comments that we 
> want to have meaning on (my embarass case) and comments we don't 
> (Ian's preferred solution).  One solution is that you only use 
> rdfs:comments for those things that you want to have meaning on, and 
> use xml comments on things that are illustrative.  Ian doesn't like 
> this. Ian, can you make a suggestion that can handle both contentive 
> and non-contentive comments -- just ignoring the use cases for 
> contentive comments won't make them go away

I have explained in detail why XML comments wont work. I have also
explained that you already have a mechanism for adding semantically
meaningful strings to the ontology. So can we please agree to keep
annotation strings as annotations!

Ian


>   -JH
> 
> 
> >Ian
> 
> 
> -- 
> Professor James Hendler				  hendler@cs.umd.edu
> Director, Semantic Web and Agent Technologies	  301-405-2696
> Maryland Information and Network Dynamics Lab.	  301-405-6707 (Fax)
> Univ of Maryland, College Park, MD 20742	  240-731-3822 (Cell)
> http://www.cs.umd.edu/users/hendler
> 
Received on Wednesday, 12 February 2003 17:45:48 GMT

This archive was generated by hypermail 2.2.0+W3C-0.50 : Monday, 7 December 2009 10:57:57 GMT