RE: [tm-pubsubj] Subject identification and ontological commitment : areal-world example

["Bernard Vatant" <bernard.vatant@mondeca.com>]

Mary

Was your message intended to be private or to the list? My answer is public
:)

> > Why? The dissidents could publish their own set of PSIs making
> > clear that they disagree with ISO PSIs (in OWL, they could assert that
using
> > owl:incompatibleWith). And up to the users to choose their
> > authority. But if applications use ISO PSIs, then let them be
conformant to what ISO
> > asserts.
>
> This is still not in the committee's scope. You think it in our
> scope to say whether we are or are not Pubsubj policemen or women, and
> describe that? :)

That is not a question of anybody, *us* or *them* making the police. We are
about recommended practices, that we think could avoid any semantic web (of
any kind or size, private or public, local or global) to become a semantic
mess. People consider too much that there should be exactly one PSI by
subject in the universe. This is clearly unsustainable. There will be a lot
of PSIs published by various authorities for similar subjects, that some
(publishers or users) may consider identical or not. My choice of such or
such PSI (see my example of country) should depend on my commitment or not
to the publisher's ontology. What I say is: if you want the system to work
correctly, there is a kind of consensus, I would even say trust contract,
to be established between the publisher and the users, in order for the
PSIs to be used consistently by everyone. And we should indeed recommend
that if you don't agree on the properties given by the publisher in this
subject indicator, don't use his/her PSIs. Certainly you don't speak about
the same thing. This is not making police. It's simply saying: if you don't
follow this rule, you will trigger a semantic mess, because people and
systems will believe you agree on this subject, though you are not.

> > > We are only providing a standard to represent the information in
> > > a computer.
> >
> >I disagree. The TC charter mentions explicitly that we'll deliver
> >recommendations for *definition, management and use*. I have
> proposed last
> >year to review the charter, but everybody seemed to be happy with the
> >original one. So management and process of use is not out of
> >our scope. I
> >would say that to-be adopters certainly wait mainly for that
> >before jumping in the wagon.
>
> This is in the context of OASIS (electronic commerce) and this is what I
> meant.
> I think we are interpreting the charter differently.
> *definition, management and use* is always in the scope of information
> systems. I really do not understand you here.

I see pretty well that we disagree on the interpretation of the charter. I
believe I always had in mind *mainly* the human and social aspects of the
three objectives : definition, management and use. Maybe I would agree the
scope is information systems, if an information system includes its human
users. Otherwise it does not make sense to me. What will users be able to
achieve with the PSI toolkit? What will they be interested in? Gather piles
of contradictory statements about what they believe (maybe misled) to be
the same subject? or build consistent, effective systems based on clear
agreement on ontology, identifiers and identification process (like isbn.nu
example)? The former is an interesting academic exercise, but the latter is
really what we are about and what is the demand now, everywhere.

So we can't be agnostic about human process, ontological commitment,
community of users, and the like. We are at a core of a very difficult
problem in which the ratio human/technical is certainly over 80/20.

Bernard