W3C home > Mailing lists > Public > www-webont-wg@w3.org > September 2002

Re: LANG?/SEM?: using resources (was Re: LANG: owl:ontology)

From: Dan Connolly <connolly@w3.org>
Date: 18 Sep 2002 10:33:53 -0500
To: "Peter F. "Patel-Schneider <pfps@research.bell-labs.com>
Cc: welty@us.ibm.com, www-webont-wg@w3.org
Message-Id: <1032363234.2992.4075.camel@dirk>

On Wed, 2002-09-18 at 09:40, Peter F. Patel-Schneider wrote:
> 
> As far as I can see, there are several ``strange attractors'' in this
> space:
> 
> Attractor 1:  Resources have definitions.
> 
> Attractor 2:  Whenever a resource is used, the intent of the ``definer'' of
>   that symbol is automatically a consequence of that use, whether that
>   intent is in the form of a definition or even in some other form.

Try this refinement: whenever a resource is used *as a property*,
the intent of the definer is a consequence.

I'm still not sure how to formalize this... my current preference
is not to really formalize it, but to make it the job of
whoever invokes the tools to tell the tool how many premises
to consult... meanwhile, I think it does address your concern:

> These two seem to be attractive at first glance, but have untenable
> consequences in both the RDF and OWL world view.
> 
> Attractor 1 has a technical problem.  There is no reasonable way in RDF to
> separate out the triples that could form the definition of a symbol, even
> if one might argue that the symbol has a definition.  Similarly, OWL has no
> notion of what would make up the definition of a class, a property, or an
> individual.  I don't see any tenable way of adding the notion of a
> definition to OWL, particularly OWL written in RDF.
> 
> Because resources can't have definitions in RDF or OWL, Attractor 2 can't
> use the notion of a resource.  However, even if there was a way of
> selecting the definition of a resource, automatically having the
> definition of a resource being a consequence of the use of the resource is
> a bad idea.
> 
> For example, suppose I want to be able to mention the Taliban in an OWL
> ontology.

If you just want to mention it, ala, "I disagree with the Taliban",
you haven't used any properties from the taliban.

But if you say
	:Bob taliban:supports :Jim.

then you mean that Bob supports Jim, where supports is used in
the sense taken from the taliban.

Operationally, this suggests that folks should feed the taliban,
as well as thi Bob/Jim document, into their tool when looking
for conclusions.

As I say, I'm not sure how to formalize it... perhaps there
are two kinds of entailment: (a) entailments that can
be justified from the document alone, and (b) entailments
that can be justified from the document plus all the
documents it invokes by use of properties. The latter
involves consulting the state of the web, which invokes
all the tricky trust and distributed computing issues.

Note: the defintion of the rdf:type property is such that
using rdf:type with a class from ontology X has the same
effect as using a property from ontology X. So if you
say
	:Bob rdf:type taliban:GoodGuy.
then you've licensed conclusions that come from the taliban.

>  I have to pick a well-known resource that is commonly used to
> refer to the Taliban.  If Attractor 2 is in force, as soon as I do so, I am
> committed to whatever the definition of that resource is.  In the case of
> the Taliban, I would be forced to choose among the various definitions of
> the Taliban, most of which would force rather strong consequences.  The
> situation would be even worse if I wanted to related the various
> definitions, as I would end up with a contradition just by making a
> co-reference between the various definition.
> 
> The situation is even worse with Attractor 2 in the absence of
> definitions.   To get the intent of the definer of the resource, I would
> have to commit to an entire document (or ontology, if you prefer).  Here I
> would end up with things like committing to the invasion of Iraq just by
> mentioning George W. Bush.
> 
> 
> As far as I can see the only viable route is to be able to use resources
> without committing to anything related to that resource.  To commit to
> something in some other ontology/document, use imports.
> 
> 
> Peter F. Patel-Schneider
> Bell Labs Research
-- 
Dan Connolly, W3C http://www.w3.org/People/Connolly/
Received on Wednesday, 18 September 2002 11:33:54 GMT

This archive was generated by hypermail 2.2.0+W3C-0.50 : Monday, 7 December 2009 10:57:52 GMT