OASIS Mailing List ArchivesView the OASIS mailing list archive below
or browse/search using MarkMail.


Help: OASIS Mailing Lists Help | MarkMail Help

tm-pubsubj message

[Date Prev] | [Thread Prev] | [Thread Next] | [Date Next] -- [Date Index] | [Thread Index] | [List Home]

Subject: Part 1 - Technical answer - RE: [tm-pubsubj] Declare victory and retire?

Lars Marius

There are two kind of issues in our exchange that I would like to address
in two different threads. This first answer will deal with technical
aspects only, and the second with TC process aspects.

> | My personal hunch right now is that mainstream technologies will be
> | tomorrow, and that means *pretty soon*, the Semantic Web package
> | ontologies-OWL-RDF. I'm more and more convinced OWL objects
> | (classes, individuals and properties) can achieve most of the tasks
> | PSIs were intended to achieve for semantic interoperability, with
> | the advantage of clean semantics.

> Uh, what?!? Bernard, are you absolutely sure that you are not here
> confusing the identification of subjects with making assertions about
> those subjects?

It's not a confusion, or if it is, it is deliberate, based on the firm
conviction that those things can't be addressed separately:

- We all agree that subject identification needs two things, the subject
identifier and the subject indicator. Deliverable 1 is all about that, and
how those two things are two faces of the same Janus.

- But we've not yet settled *at all* the issue of the subject indicator
content, and in fact every time we've got to that hard point, we have
stumbled on it and skip to some other issue. To indicate the subject, you
have to make some assertion(s) about it. If you don't make any assertions
at all, you have a kind of empty subject indicator, and a subject
identifier that nobody can makes sense of. What use do you have of
http://psi.oasis-open.org/geolang/iso639/#dan is the matching subject
indicator does not contain some minimal assertions, like the fact that the
subject is an individual language, IOW an instance of the class defined
itself by http://psi.oasis-open.org/geolang/iso639/#language.

We have been around and around it. If I define (or use) a PSI for the class
"Malus Domestica", does it make sense if I do not commit to any
class-subclass schema this class is explicitly declared as participating
in? I might have been not clear about it in the past, but now I think I am
: it does not make sense, or it opens the door to such ambiguities that the
subject indicator can indicate as many subjects as you want. If I want to
explicit this subject, I will declare that it is a subclass of "Rosaceae",
itself a subclass of ... an the best way to declare without ambiguity this
ontological commitment is to have the subject indicator as a formal class
or individual in a formal ontology. There enters OWL ...

That is the sense of what I propose at http://www.mondeca.com/owl/lang.rdf.
As explained to Lars Marius in private exchange, this is not a file
containing just RDF metadata about the GeoLang PSI set, it is a proper
standalone ontology of which every class and individual is intended to be
used as a subject indicator. For interoperability sake only, those subject
indicators are declared equivalent to those defined by GeoLang, using
"sameAs" declaration. But that was optional.

> OWL does not do anything about the first; it is all
> about the second. In other words: OWL and pubsubj are orthogonal.

Honest, I don't understand how you can say that OWL does not do anything
about identification!
What is "http://www.mondeca.com/owl/lang.rdf#Lang" if not an identifier of
a class, hence a built-insubject identifier?

OWL is about identification and assertions about subjects that are
identified. I don't see orthogonality at all, but 90% overlapping. Unless
we are not at all on the same page from the beginning (which may be the
case after all).

> Frankly, what you write here reminds me of the early story in
> ComputerWorld that discussed whether XML or Java would win. Obviously,
> the whole thing was a non-issue.

I don't know enough of that story to catch the analogy, but obviously a
programming language and a markup language are not about the same tasks.

> Pubsubj works equally well for RDF and topic maps. You can use
> published subjects with OWL. So what is the problem?

The problem is that people are about to use OWL classes and individuals as
"de facto" subject identifiers, and that I don't see how and when we are to
catch up with that, what kind of content we will propose for subject
indicators that OWL elements will not achieve better, with more clarity and
less ambiguity.

But that goes along with process issues I will address in Part 2.

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

[Date Prev] | [Thread Prev] | [Thread Next] | [Date Next] -- [Date Index] | [Thread Index] | [List Home]