Using ontologies to compare resources was: Re: The range of the HTTP dereference function

Joshua Allen wrote:

(note this is a comment, not a proposal)

> > From: Jonathan Borden [mailto:jonathan@openhealth.org]
...
> > http://www.markbaker.ca/ can easily both identify "Mark Baker", the
> person,
> > and be GETable.
> >
> > The media type image/jpg _representation_ might be a picture of Mark.
> > The media type text/html _representation_ might be an HTML page which
> Mark
> > uses to describe himself.
> >
>
> Absolutely!  You are agreeing that we need a way to clarify in which
> "sense" the URI is being used.  If I store a triple that says,
> "http://www.markbaker.ca/ looks-like http://www.joethomas.fr/", this is
> meaningless unless I specify that whether I mean the "text/html" or the
> "image/jpg" associated with that URI.

This is the (in)famous resource-representation bug.

The resource is the central node to which various properties are attached.
One can think of the various resource representations as such properties.
Triples may add other properties. What you are comparing above is not two
resources, directly, rather the visual appearance of two resources. Consider
this a property named "visual-appearance" or "image/jpg" (either works). So,
" http://www.markbaker.ca looks-like http://www.joethomas.fr " is (I
presume) intended to mean:

visual-appearance-of(http://www.markbaker.ca) looks-like
visual-appearance-of(http://www.joethomas.fr) and it works.

>
> We need to allow that distinction to be made, and we need to allow it to
> be made without requiring the resource or it's representation to be
> online at the time it is made (or consumed).
>

This is one of the things that the Web Ontology language will do, and that
DAML+OIL does today. You can say that certain properties are, for example,
uniquely identifying, such that if the values of such properties are equal
then the resources themselves are equal. For example, a US based healthcare
ontology might state that the social security number is a unique identifier
(although it isn't but close enough to use as an example :-)

We sometimes do take a photo to place in a patient's chart as an identifier.
You could presumably "merge" two charts together without the patient being
there. As these tricky operations are being spread across a WWW, ontologies
etc. become very useful.

Jonathan

Received on Thursday, 28 March 2002 00:00:08 UTC