RE: SUO: Re: SUO Ballot with 2 Questions
Dear Jean-Luc,
All of a sudden I am being quoted a lot - very worrying.
Mostly I agree with you, but see some comments below.
Matthew West
Principal Consultant
Shell Information Technology International Limited
Shell Centre, London SE1 7NA, United Kingdom
Tel: +44 20 7934 4490 Other Tel: +44 7796 336538
Email: matthew.west@shell.com
Internet: http://www.shell.com
> -----Original Message-----
> From: Jean-Luc Delatre [mailto:jld@club-internet.fr]
> Sent: 05 June 2003 12:26
> To: Mike Pool
> Cc: Patrick Cassidy; standard-upper-ontology@ieee.org;
> semanticweb@yahoogroups.com; cg@cs.uah.edu
> Subject: Re: SUO: Re: SUO Ballot with 2 Questions
>
>
>
>
> Mike Pool wrote in
> "http://suo.ieee.org/email/msg09573.html";
>
> > Even, if the case can be made that there will never be one
> and only one top
> > level ontology, which seems odd to me if we make the
> ontology sufficiently modular and extensible,
>
> Okay, okay, I will try once more to explain why it is
> IMPOSSIBLE to have
> a single ontology (top or NOT)
> as soon as two separate entities (groups or individuals, no
> matter) claim to rely on it.
> Along ongoing discussions on SUO and elsewhere the rationale
> for this has been burried in a flurry
> of other arguments,
> which, if relevant too, are not as crippling to the SUO
> purpose as this very basic one.
>
> First, reminding a few definitions.
>
> Any concept in an ontology refers to both a set of
> individuals (dogs, tables, ...)
> and a set of attributes (having 4 legs, barking, ...).
>
> The first "meaning" is a what is known as a definition by
> extension (named as such since Aristotle
> up to Peirce and nowadays).
> The second "meaning" is a what is known as a definition by
> comprehension (same remark).
> Both must coincide, that is, all individual dogs meant to be
> mentioned by some ontology
> should have ALL the attributes ascertained for dogs by this
> same ontology.
> Conversely, all individuals that show all the attributes
> ascertained for dogs MUST be dogs.
>
> Now, while it's obvious that everyday we agree on some
> partial "chunks" of ontology, provided we
> don't dig "too deeply",
> why is it that two people CANNOT agree on the same OVERALL ontology,
> that is, an ontology that would cover ALL their concepts?
>
> Because, no two people or groups share ALL theirs purposes and views!
> And, as Matthew West has put it in:
>
> http://suo.ieee.org/email/msg09581.html
>
> MW>I personally think we would end up with more than one ontology,
> MW>but for GOOD PRAGMATIC REASONS - there are applications
> that would favour one style over another.
>
> (my emphasis on "good pragmatic reasons")
MW: I am quite happy with the emphasis.
>
> So what is the effect of these "good pragmatic reasons" that
> brings TROUBLE
> and prevents the merging or even just the INTEROPERATION of
> any two ontologies?
MW: Preventing interoperation is too strong, but achieveing it
certainly requires considerable effort - at least at present.
>
> On one hand, for a given concept, assuming it names the same
> "extension",
> i.e. the same set of individuals, in both ontologies,
> many times the "interesting" attributes are not the same ones!
>
> But this is easily delt with, the second problem IS the crippling one:
>
> Due to different views and purposes, it happens frequently
> that two NEARLY SIMILAR concepts from two
> distinct
> ontologies name, in fact, DIFFERENT sets of individuals!!!
MW: Indeed. An extreme example of this is 3D vs 4D where concrete
objects are distinctly different in nature (being in one case extended
in time and in the other case not).
>
> This is where all hell breaks loose.
> Assume for instance the common confusion between whales and fish,
> where for ontology A anything that lives in water and is
> edible is fish,
> and for ontology B fish is the usual zoological definition.
> How can B talk to A about whales, and how can A talk to B
> about, say hippocampus, not really edible
> and still a fish?
>
> Neither has the concept that would allow to name the
> individuals intended by the other party.
> What happens when both parties are humans is that we embark
> in explanations,
> we upgrade the OTHER people ontology in order to convey our
> proper meaning.
>
> COMPUTERS DON'T DO THAT!
MW: Yes, but the whole point of the proposal here is to allow
this to happen (mediated by people at least in the first place).
>
> Furthermore, this on-the-fly upgrading of ontologies is
> rejected in principle
> for the sake of "stability of meanings",
> actually more for the fact that current programming practice
> is UNABLE to deal with such "forking"
> semantics.
MW: That is a bold statement that I do not think stands up to
scrutiny.
MW: In your example above the original concepts of fish (say
fish1 and fish2) would be retained with information about who
used them when. If an improved and more widely shared concept
arose, it would be added - say fish3 - but the original concepts
would be retained, but perhaps put on a list of "historical"
rather than "current" concepts. You could even track the
relationship and development. I don't see anything here that
shoudl defeat curretn programming practice, at least not that
with which I am familiar.
>
> It could be argued that the more refined the concepts
> subdivisions are the closer we get to an
> "universal ontology" where
> any possible subset of individuals of interest to someone
> could have a proper name, but alas this is
> an
> infinite process that would create a huge and ever evolving
> ontology, of unmanageable size
> whith only a tiny subset of the whole stuff being of use and
> interest for any given user.
MW: This true already with langauge today. Moast of us only use
a small subset of it - and then in a particular way. The key is
being able to understand and share the small subset that is needed
for some purpose, and be able to refer to it unambiguously.
>
> Quoting Matthew West again from the same message:
>
> MW>They lack hundreds and thousands of manyears of work which
> is probably
> MW>required to achieve maturity.
>
> How big will these ontologies be after those thousands of
> manyears of "maturing"?
MW: Actually I had in mind better rather than bigger, but they would
be bigger too - but by then not just upper ontologies. My issue is
that current ontologies are immature because of the poor way they
map to the "real world". I would expect more time to give more accurate
ontologies as much as larger ones, and it is our ability to accurately
reflect the "real world" which is my greatest concern.
>
> And what about Seth Russell's concern in:
>
> http://suo.ieee.org/email/msg09583.html
>
> SR>I still must choose to rely on either a proprietory
> corporation's efforts or a industry
> consortium.
> SR>It is a difficult choice. Which scheme of things will be
> around in 5 years, in 20 years ?
> SR>On the other hand if the IEE can establish methodology
> such that my data can be translated
> SR>into whatever my target audience is using at the moment,
> SR>it doesn't really matter which one I choose.
MW: In making a choice of course for some purpose, you potentially
isolate yourself from those that have made a different choice.
>
> Then, what about a lattice of ontologies, a registry, motion
> 2 or even dynamic interoperability of
> ontologies?
MW: Not sure what your point is here.
>
> Best regards.
>
> -- Jean-Luc Delatre
> --------------------------------------------------------------
> ---------
> "Always do right - this will gratify some and astonish the rest."
> -- Mark Twain (1835-1910)
> --------------------------------------------------------------
> ---------
> http://perso.club-internet.fr/jld/ -- GSM: +33 6 11 24 06 29
>