RE: subject indicators ... ?

Stella, Alistair, Charles, and all

> If I've understood correctly, the idea is to point (using a URL) to a
> place where a definition of the concept appears.

Stella, you understood *almost* correctly :)

In fact, the creators of the notion of subject indicator were not as ambitious and
arrogant as to figure that subjects can always be *defined* absolutely. Remind you that in
the original context (topic maps), a subject is *whatever can be talked about*, not only
concepts in structured concept schemes. Providing a proper and absolute definition of a
subject might often be difficult or even impossible, but a well chosen resource can
"indicate" by any means what the subject is about. In a scheme containing color concepts,
e.g. "Lavender Blue", the subject indicator resource could be a sample of this color,
along with its RVB code.

> (And this is not the same as pointing to a place where an example appears)

Well, for some subjects, a bunch of examples could be more efficient to indicate the
subject than a convoluted definition :)

> Rather than pointing to that
> other place, is it not better just to give the text of the definition?

All the point of subject indicators external to the scheme is to enable identification of
identical concepts in different schemes, whatever the process this can trigger, like
merging of topics in topic maps, which was the primary intended use of subject indicators.

But actually, I figure that a subject indicator for a concept in scheme A could be a
concept in scheme B, itself providing a definition that A doesn't care to reproduce. In
fact I am curious to know what folks think about the use of subject indicators vs scheme
mapping.

> There could be an advantage if the text is too long or complex to
> import; or if the aim is to allow the definition to vary according to
> the dictat of the owner of the resource pointed to.

Well, certainly, using a subject indicator that you don't control is some kind of risk,
and means common trust. That's why was introduced the notion of Published Subject
Indicators, that should include all the context necessary to support a trustable use of
subject indicators.

For example http://antwrp.gsfc.nasa.gov/apod/astropix.html is a quite good subject
indicator for the subject "Astronomy Picture of the Day" : "Each day a different image or
photograph of our fascinating universe is featured, along with a brief explanation written
by a professional astronomer."

This resource is changing every day according to the declared dictat of its owner ... but
is in fact remarquably trustable. It's been there since '95 and is one of the most stable
changing resource I know of on the Web ...

> For the sake of the integrity of the original scheme, it would seem safer to be sure
what
> the concept is and not leave it to someone else's control! ( That way
> you can be more confident of the relationships with other concepts in
> the same scheme.)

Well, this has been debated over and over in various Semantic Web lists. All the idea of
the "Web of Trust" is indeed to have  resources defining, and authorities publishing,
concepts, and users using them in full trust, the same way you trust the network when you
send an email or put an URL in your browser, and assume it will get to the right person or
retrieve the relevant resource. You have the same issue with the OWL use of "import". When
you import an external ontology in your own one, you trust the imported source, either to
be stable, or to change in a way that will not screw up your own.

In fact, this is a fundamental choice in open environments, and experience indeed shows
that it does not work that bad. Even if you have a breakdown of trust in 10% of cases (and
this is very pessimistic, the actual figure will certainly be lower), what you gain in the
other 90% is really worth it.

Cheers

Bernard

Bernard Vatant
Senior Consultant
Knowledge Engineering
Mondeca - www.mondeca.com
bernard.vatant@mondeca.com

Received on Monday, 27 September 2004 07:34:42 UTC