Re: SUO: *Date 19 Mar 2002 -- Modus Ponens
Sorry,
It's seems I have kicked the ant's nest with my message:
http://suo.ieee.org/email/msg08117.html
That's not what I intended, specially not to revive the fight
between John Sowa and Adam Pease.
Trying to settle down this and not to spread it into too many messages,
I will here answer messages from,
DM - Douglas McDavid: http://suo.ieee.org/email/msg08118.html
CM - Chris Menzel: http://suo.ieee.org/email/msg08119.html
JS - John Sowa: http://suo.ieee.org/email/msg08120.html
AP - Adam Pease: http://suo.ieee.org/email/msg08123.html / msg08126.html
BA - Bill Andersen: http://suo.ieee.org/email/msg08125.html
May be not always in that order.
I will use JD to disambiguate some of my own quotations.
Anyway, *my* position is not really close to any of yours
and neither fully understood, I will try to clarify that too.
DM> Some interesting ontological categories you seem to have in mind, Jean-Luc:
> "SUO crowd"
> "Bateson bunch"
Loosely worded and probably not proper concepts (not closed under |><|)?
Yes, but everyone of us use such denominations everyday without much harm,
because we tacitly add more restrictions to the explicit wording and even
shift the meaning according to the context of the utterance.
Different locutors make different assumptions and this is where
misunderstandings happen, but usually not so often and not so severe.
And I bet few on the list had a hard time finding a meaning to my words.
But the story goes another way when we "talk" to computers!
"SUO crowd" could mean "all subscribers to the SUO list"
or "SUO crowd" could mean "all subscribers to the SUO list" minus me
or "SUO crowd" could mean "all subscribers to the SUO list" minus me, Jon and whoever else
or "SUO crowd" could mean "all true believers in the SUO Scope & Purpose"
etc...
If we have to enter "SUO crowd" into some knowledge base which concept
will the computer choose among those above?
It *has* to depend on where the words "SUO crowd" come from.
The algorithms will never be as clever as us to figure out the context
(altough bayesian inference from surrounding words can do quite well),
but at least the *possibility* of mapping the same lexemes to different
concepts should not be excluded by the very structure of the ontology.
That's what I have in mind when I argue against "*true* denomination"
or when I say:
JD> That is, if you need not just to rename a concept, but to swap part of
> it's definition for a stronger or weaker one you cannot.
> See my argument about 'penguins' in my message
> http://suo.ieee.org/email/msg08062.html
> If it ever happens that a concept has to be weakened or strengthened in
> the top ontology the modification of existing applications will be much more
> expensive than without this flexibility.
There must be an extra layer of mapping between the lexemes and the
concepts and it must be part of the ontology structure.
I may be wrong but I don't see this anywhere right now.
I have finer points to make about that further down on this mail.
DM> I suppose you imagine a "Delatre cadre"? ...
Exactly!
and a "McDavid cadre" and a "Awbrey cadre" and a "whoever cadre" etc...
In order to meaningfully relate statements expressed from within
different such "cadres" I do not expect any *top* or *upper merged*
preset ontology to be up to the task.
In each case of a connection to be made between two statements
from different cadres a specific check has to be made to adjust
the concepts referred to by each statement.
This, of course, will require the use of some already existing
agreement upon semantics from the different cadres,
a "common ontology" I grant you that.
But, unless those statements belong to a domain of discourse
which has already been crisscrossed at length by the two parties,
it is very likely that the finer tuning bewteen the semantics
cannot possibly be done with already available concepts.
In many cases you will have to setup more refined concepts
from the existing ones and the statements contents in order
precisely match their meanings.
***********************************************************
* *This* is what has to be automated, don't rely on *
* a frozen ontology however "top" it might be! *
***********************************************************
Only the semantics involved in a given exchange need to be setup.
Fortunately, when learning English, you don't have to memorize
the whole OED before you start speaking.
DM> ... As Carl Sandburg implies,
> everything starts with a dream.
Having the dream come true requires more than just dreaming.
The dream has to be expanded *and* trimmed, not too much of each.
JD> > Some of the "assumptions" that, to me, seem questionable and shared
> > by most people involved in ontology projects (alas, not only SUO!) are:
> >
> > 1) There exist ultimately a *perfect* all encompassing ontology that allows
> > describing everything, and we have to chase for it however distant it be.
CM> I certainly agree that that would be a dubious assumption, but it seems to
> me that the prevailing view among participants on this project is that (a)
> there is no all-encompassing ontology, maximally expressive ontology and
> that (b) there there could be multiple ontologies -- even "upper level"
> ontologies -- that reflect different ways of conceptualizing the world.
> Exactly what evidence do you have that anyone involved in this project
> actually accepts the above assumption?
See John's message...
http://suo.ieee.org/email/msg08120.html
JS> I suggest that you look at what Adam has been proposing: a single
> monolithic standard, named SUMO, which he hopes that the IEEE will
> endorse.
Just teasing, I'd rather not, but it's so...
I would put it another way.
As for (a), even if the "prevailing view" among participants is that
there is no all-encompassing ontology, they do not seem to have drawn
all the conclusions and consequences from this simple fact.
To me this entails that the "top" ontology has to be almost void
because no stable, general enough, ontology will fit a majority of
practical cases and will be agreed upon by the participants.
********************************************************************
* It is the *match and merge protocols* between two separately *
* designed ontologies which have to be agreed upon. *
********************************************************************
As for (b), departing from John's position, I find this irrelevant.
It does not matter if the "top" ontology is monolithic, multiple,
modular, layered, latticed (?), whatever.
There will be "infinitely many" (JS) ontologies and all we have to
care about is not even to merge any two of them but to reconcile
some concepts from both *where it matters for the task at hand*!
JD> > 2) There *must* be for each concept a *true* denomination ...
<remember the context, I not asserting that>
CM> Now that is completely bizarre. There are a number of legitimate concerns
> about the SUO project that can be, and indeed have been, raised, but that
> is surely not one of them. No one involved in this project (or any
> ontology project that I know of) thinks that the names of classes,
> properties, relations, etc are anything but conventional. Any push for
> one name rather than another is motivated only by pragmatic concerns, such
> as the avoidance of names with natural language connotations that might
> hinder a user's understanding of the ontology.
I agree that the names should be conventional and this is unavoidable for
the *nearly void* top ontology I suggest and I even go farther:
Apart from a very small number of thorougly screened basic names from
this ontology *no* other entities matching from two different ontologies
should rely on identity *or even one-to-one matching* of names.
Do *not* use a mapping from "chicken" to "poulet" (in french, for instance)
to relate concepts from differents ontologies.
And this even for mathematical concepts, for, as you know, different
authors often disagree on terminology. See next remark for more.
JD > > and everybody have to agree to use it.
<remember the context, I not asserting that>
CM> Well, what would you suggest to people who want to use the concepts in a
> given ontology? That they choose their own terms for those concepts and
> then write programs that translate the axioms of the ontology into
> identical axioms that simply use their own lingo? How stupid would that
> be?
This is not stupid, but, alas, unavoidable!
Because, tough two different concepts may appear *nearly* the same, and
just in need of a "sensible" renaming, their meanings may not exactly
coincide if coming from different ontologies.
And even from the *same* ontology! Synonyms do *not* have truly the
same meaning, you can ignore that most of the time, but in some cases
it matters.
Those cases, *if* they are to be properly handled by the application
using the ontology, will require differences in the internal
representation of the given synonyms.
CM> If you've got credible objections to the SUO that you can document, I'm
> sure the SUO participants will be happy to debate them. These objections
> are not remotely credible.
Some misunderstangings may be?
I hope I make myself clearer in the present mail.
JS> Chris and Jean-Luc,
[snip]
> I believe that they are very credible as objections to SUMO. And some
> people, Adam P. especially, have claimed that SUMO conforms to the SUO
> charter, but other proposals such as IFF and my lattice of theories
> do not.
Thanks John.
You can ease a bit on Adam now.
I am probably pissing him of more than you do by then.
AP> I've said this before, but I do believe that there are many possible upper
> level ontologies. We're trying to create one that is good enough.
Good enough for what?
AP> I disagree with (a) and agree with (b).
See above my replies to Chris on (a) and (b).
AP> The fact that there may be many
> possible upper ontologies is not an argument against agreeing on
> one.
Oh, yes it is!
Unless you trim it down to the barest you can,
but you are just going the opposite way.
AP> Standards exist in fields whether there are many possible solutions,
> but the value of interoperability, common tools, etc leads people to adopt
> a common standard.
I don't think any other fields have such deep involvements not only in
philosophy, metaphysics, etc.. but more seriously with ultimate impact
on feasability of future developments in the whole software industry.
If you screw it that will cost *billions* to future users.
Reminding you of Pat Hayes:
http://grouper.ieee.org/groups/suo/email/msg03801.html
AP> True but it's unclear what the "fundamental design" would consist
> of. We've had several years on this project with most people just talking
> and not proposing any standards words. That sort of effort doesn't lead us
> to a standard.
Obviously, there must be something wrong somewhere!
Whom should I quote?
Seneca: If one does not know to which port one is sailing,
no wind is favorable.
I mean, can you define where you want to go with more precise
words than a "good enough top level ontology"?
Or shall I quote Marcel Schoppers again:
(http://groups.google.com/groups?selm=4187%40uiucdcs.UUCP&oe=ISO-8859-1)
> If AI has made little obvious progress it may be because we are
> too busy trying to produce useful systems
> before we know how they should work.
Can you make any other suggestion than pushing forward?
JS> >Unfortunately, Teknowledge is the only company involved with
> >the SUO that has enough funding to have people working full
> >time on developing axioms. For me, SUO is a part-time effort,
> >and I am working on much more promising endeavors than fighting
> >an uphill battle to make improvements to SUMO.
AP> The battle need not be uphill, nor need it require any full time
> commitment. If everyone on the list expended effort in writing axioms
> either expanding SUMO or in proposing a concrete set of axioms embodying an
> alternate theory, we'd make tremendous progress.
I am myself lazy and not always too smart, so I understand very well that
you would welcome some outer contributions to advance your work.
> If you would take the
> time to do that instead of writing messages that are just references,
> discussion or opinion, you'd have time for a substantial contribution.
Making sound contributions is *not at all* the same degree and extent of
work than making comments, even well argumented, and as John said,
JS> Teknowledge is the only company involved with the SUO that has enough
> funding to have people working full time
and that does make a *big* difference!
JS> >You keep missing the point. There are infinitely many
> >upper-, lower-, and middle-level ontologies. And no single
> >one can ever be best or even adequate for all purposes.
AP> I don't believe I've missed your point. We just disagree. You've asserted
> your point but haven't proven it. You're welcome to your opinion. But
> participating in an effort to create a standard ontology only to say that
> the objective is wrong is counterproductive.
Hmmm... May be not, be carefull, watch out for a Darwin Award.
You've got some budget and you are perfectly entitled to burn it
in any way you see fit.
But how would you feel if shortly after SUO/SUMO is completed, or even
before, the practical and commercial value of it is closer to "zilch"
than to anything else because some other technology allows to build up
customized ontologies about almost anything at a small fraction of the
cost of developping SUO/SUMO or even the cost of installing it (including
users training of course)?
JD> > Excellent Jon!
> > Are you joining me against the SUO crowd?
BA> Is this important to you? To Jon? To Chris Lofting?
To me?
Yes, *billions* (for compagnies) and hassles (for people like me) at stake.
See above.
BA> If so, I again extend
> my invitation to you and all like you stop talking and fire up a research
> project of your own and build something. Then let the rest of us infidels
> run the code on some real world problems. We'll let you know if it works.
Budget. Just see above, John is with me on that too.
BA> Well, I'm assuming you object to the search for the "perfect" ontology that
> allows describing everything on the grounds that it's bad philosophy. That
> in turn leads me to believe you have a counter-philosophy. And in this you
> do not disappoint:
I am not arguing here about "philosophy" but about practicality:
assuming that there is somewhere an "existing" if however unreachable
perfect goal makes you unable to settle down for "realistic" goals.
The runaway can restart at any time you encounter a problem, even a minor one.
I guess this might be what undermines the quest of people like Jon Awbrey,
but that's just a guess, he may not agree.
JD> > 2) There *must* be for each concept a *true* denomination and everybody
> > have to agree to use it.
BA> You (along with all other anti-realists) assume that what is being talked
> about when building ontologies is "concepts" (your word, not mine). In this
> you deny that there is anything real at all - the only thing that counts is
> how we talk about it.
NO, the only thing that counts is: does this approach allow us to reach our
arbitrarily choosen practical goals more effectively?
As my wording on the previous remark may suggest, I consider myself more
of a "realist" than people that insist on "real objects".
Whatever that may mean I consider it a *void* metaphysical questioning.
BA> You have taken a position, whether you realize it or
> not, which goes far beyond what the poor chaps working away on this list to
> produce something which might or might not work are doing. It is you who
> are obsessed with philosophical questions and not them.
I am not "obsessed", unfortunately for the "poor chaps", in this complicated
matter of ontology building, taking firm positions with respect to
"philosophical" choices is key to successfull *design* choices.
BA> So, since you are so willing and adept at making "speedy" (your word, not
> mine) judgments about what people are and are not doing in this mailing
> list,
Could it be that my judgments are not so "speedy" nor "impetuous" than
they appear?
But don't take it hard. ;-)
BA> allow me make one of my own:
>
> Some of the "assumptions" that, to me, seem questionable and shared by most
> people involved in CRITIQUING ontology projects are:
>
> 1) If one doesn't like what some other person has to say about a subject,
> then that other person's position is labeled as bad philosophy, or "deeply
> ingrained" prejudice.
This can be debated, it has to be stated first.
I know critique always hurt american people more that people from other
cultures because it contradicts their optimistic view of the world.
I cannot help but state my opinions and *then* argue about them.
Long ago, an american software guy named Wayne Rice complained against
me too and said that I was too often judging that things were
"Positively no good", would that wording be more acceptable to you? ;-)
BA> 2) One's own position is simply the TRUTH.
Not *the* truth, but what can be demonstrated as better under
the current conditions (modulo misunderstandings...),
that's what arguing is for.
JD> > There is also a funny anecdote (reported by Bateson or someone of
> > his bunch, don't remember) about the german speaking inhabitants
> > of Tyrol getting angry about italian speakers calling a horse
> > "cavallo" where it's obvious the right name is "pferd".
BA> No, the right name is "Pferd". After all, we're talking about names (your
> word, not mine) and not "concepts". Versuch's noch mal.
The very reason that I object about mapping lexemes directly to concepts,
"pferd" and "Pferd" must obviously refer to the same concept in this case.
But, wait, where does your "Pferd" name comes from?
A movie? the proper name of one of the characters!
Ah! May be we should map that to some more restricted concept than "horse".
See?
------------------------------------------------------------------------
More on the extra layer of mapping between the lexemes and the concepts.
Assuming that you have to match some statements using an english
sourced ontology versus a french sourced ontology.
The translation of "big" is "gros", but not always!
Their proper range of usage overlap but do not coincide.
Some uses of "big" can only be rendered by "gras" (fat).
While some other uses can only be rendered by "grand" (great/grand).
Conversely while "grand" (in french) usually means "tall"
in other cases it means "great" (as above) or "grand" (english meaning).
There is no hope of finding a proper *direct* mapping, the range
of meanings (nodes in the FCA lattice) has to be subdivided in
finer grains such as to be able to map each of "big", "fat", "tall",
"grand", "great", "gros", "gras", etc... to the proper node.
And this, *possibly while having to rely on context* to choose
among possible bindings. This means part of the logic will have
to be outside the ontology proper, thru some call back mechanisms.
And there we are, already drowning in the implementation details,
before we even finished dealing with the design as such.
But what *do* pertains to the design is the fact that we do not
need to care about explicitly providing nodes in the FCA lattice
unless we we meet the ambiguity.
We certainly should *not* try to cover every such possible case
in advance in a "merged" ontology nor have to rely on human
intervention and judgment when a case arise.
Therefore, what need to be defined are the brokerage protocols
to be used to automatically reconcile the semantics of cases
like "big" versus "gros" and what kind of extra informations
are needed to do that.
Will that do?
I feel my contributions to this list, specially the present one,
to be *valuable* informations about the design, if not of a "top"
ontology, but of interoperability of ontologies.
Not "just references, discussion or opinion" (AP),
too bad the SUO Scope & Purpose assumed from the start that
such interoperability must rely *mainly* on a "common ontology"!
Cheers.
-- Jean-Luc Delatre
-----------------------------------------------------------------------------
The early bird catches the worm, but the second mouse gets the cheese.
-----------------------------------------------------------------------------
http://perso.club-internet.fr/jld/ -- GSM: +33 6 11 24 06 29