Tuesday, December 20, 2005
Lattice of ontologies
Function/structure descriptions
Ian,
The issue I addressed was the need to !! sometimes !! not
have a specific URI as referent to an individual, or to a concept.
At core there are some basic limitations that come from
the RDF/W3C paradigm where terms, to be proper, need one precise
definition. This requirement is for the
purpose of equipping terminology with first order logics. However, linguistics and others have long
argued that terminology cannot be universally treated as the elements of a
logic.
So the RDF/W3C standards create - if successful - a
semantic web defined by computer scientists and imposed on user
communities. If you want something done
you better make sure that it is allowed.
Business and government have real problems with the information system
that is in place does not recognize the real time realities.
This limitation is related to the question of an anonymous
individual. One simply wants the
individual not to have to have a fixed meaning nor a complicated URI.
The same can be said about concepts. Sometimes the power of a concept lies in its
ability to shift interpretation so that situations/individuals with differing
environments can manage to "communicate". Some people call this pragmatics, where the real inconsistencies
and non-coherence between points of view are allowed because this is the way
the social and physical world !! sometimes !! is. (sorry about this philosophical statement, but it is intended to
point out an aspect of reality.)
We know that there are more than one way to work around
the limitation. One work -round is to
manage things within and across namespaces.
Here we come across the same shell game, one is either unwilling to look
at the failures of namespaces, or one realizes that so far there has not been a
resolution to fundamental issues that arise from the W3C paradigm.
This is hard stuff, as you know Ian. But the current state of Protege as the
major editor/viewer of OWL specified ontology makes this ever harder - in many
peoples views. Just check the
discussions on the Protege - OWL forum.
Your paper, referenced below, is esoteric; and the
business government communities need something that works now. So, many are asking if W3C standards ask too
much. I ask if the W3C ask the wrong
things of society.
The core difficulty, that my mind has, is with the way
that OWL and RDF files generally look with all kinds of URI references. RDF was supposed to be readable!!!! This stuff is neither readable nor
understandable - unless one dedicates considerable time and talent to the
subject - as you obviously have. The
work is very admirable. But being
admirable does not make it easy to understand, easy to use, or correct.
But then there are the performance issues and the issues
related to Protege versions and a whole lot of inference engines.
It is not my intention to be unhelpful to anyone. However, the user communities need help in
understanding why there are continuous problems with Protege and OWL.
I again hope that Protege team is able to get stability to
a project that has gone on now for 11 - 13 years. The theoretical issues may never be solved (because the issues
assume computability over common semantics).
see discussion about Robert Rosen's view on complexity
http://www.ontologystream.com/beads/nationalDebate/314.htm
So it is up to the Stanford Protege team to have a meeting
and decide if the ontological commitment to ontology only with logic is serving
the user community.
The other issue is the use of terms like
"nominal"
A nominal is a word that differs grammatically from a noun
but functions as one.
from
http://www.sil.org/linguistics/GlossaryOfLinguisticTerms/WhatIsANominal.htm
OWL connotation for "nominal" can be read from
http://en.wikipedia.org/wiki/Pellet_OWL_Reasoner
and well as your own paper
Reducing OWL Entailment to Description Logic Satisfiability
at
http://www.cs.man.ac.uk/~horrocks/Publications/download/2003/HoPa03c.pdf
which I recommend everyone read.