Vocabulary+SV

[Norman Gray]

Andrea, hello.

On Mon, 10 Sep 2007, Andrea Preite Martinez wrote:

> I think criticism is always useful, even when the way it is expressed
> goes beyond the understanding of those that are criticized. I wonder
> why the same
> criticism has not been put forward for e.g. data models, to which rdf
> is probably more appropriate than xml schemas.

First of all, I hope I haven't gone beyond anything.  I had aimed to be
emphatic (I wonder if I have been too indirect when making this argument
before), but I certainly did not intend to be uncivil, and I do apologise
if it seemed like that.

I have made the argument in DM contexts before, without, I fear,
much effect.  At the same time, I think that one should not make too
firm a distinction between DMs and vocabularies.  A vocabulary, and
a particularly rich ontology like the Ontology of Astronomical Object
Types, are at opposite ends of a single spectrum, with IVOA-style DMs
somewhere in the middle.  All of them are organised knowledge.

Tony and Kjetil mentioned SKOS as a standard, and much discussed, way of
defining vocabularies.  If you cropped the top and bottom of a SKOS file,
it's unlikely you'd know it had anything to do with RDF.

> The main motivation of our paper is "interoperability".
> Different groups talking different astronomical dialects can
> understand each other just declaring what the "equivalence" of their
> concepts is against a standard way to define those concepts. This is
> the original problem raised by the VOEvent WG when trying to fill up
> the "what" field and be sure to be understood by other communities.
> But other groups have the same problem.
> The solution we propose is a sort of astronomical "lingua franca" or
> "esperanto" (the SV + UCD).

Interoperability is of course the goal.  But where the problem
is a multiplicity of in-use vocabularies, adding a further IVOA
vocabulary is a solution which is at the very least counterintuitive.
Since the various vocabularies have grown over time to best suit their
respective domains (journals, outreach, databases), any artificial and
expensively standardised 'esperanto' will necessarily be less well suited.
Translating from a domain-specific vocabulary into an IVOA one, and then
back out again, is two hops where meaning can be lost.

As a solution, I can only suggest the utype-resolver service which I
demonstrated in Beijing.  It addresses precisely this problem, and a
complete implementation is running as a service now.  That supports
a non-centralised model of mutual explanation, without adding any
vocabularies, where everyone is free to use the vocabulary most suited
to them, simply with added interoperability.

Best wishes,

Norman


-- 
---------------------------------------------------------------------------
Norman Gray  /  http://nxg.me.uk
eurovotech.org  /  University of Leicester