RE: SUO: Maintenance - related issues
Chris,
. Comments interspersed, prefaced by: "GH2> "
.
.
.
*
*
*
*
*
*
.
.
Cheers Graham Horn
Australian Institute of Health and Welfare
================================================
Phone: 02.6244.1094
Fax: 02.6244.1199
Email: Graham.Horn@aihw.gov.au <mailto:graham.horn@aihw.gov.au>
-----Original Message-----
From: Chris Menzel [mailto:cmenzel@philebus.tamu.edu]
Sent: Wednesday, October 18, 2000 6:00 AM
To: Horn, Graham
Cc: standard-upper-ontology@ieee.org
Subject: Re: SUO: Maintenance - related issues
Graham wrote:
> > > I think John has aired some important points below, which hedge around
my
> > > stance that the benefits of the current thrusts will be minor compared
to
> > > that achievable if the SUO structure focuses on the formats of an
> > > unambiguous adaptation of natural language.
> >
> > It is good that you recognize you would need an "adaptation" of natural
> > language. To eliminate ambiguity adequately and achieve the clarity
> > necessary to ensure undertanding of an ontology -- notably, the logical
> > properties of, and relations among, its constituent concepts -- you will
> > have to regiment natural language to an extent that, in effect, your
> > "adaptaton" is a first-order language.
>
> GH> I'm quite happy with this. I believe it would benefit a large
> number of lay people to be aware that such a language would be
> available to facilitate unambiguous communication.
I don't understand, Graham. My point was that the degree of regimentation
needed to do what you seemed to be wanting would, effectively, yield a
first-order language, i.e., the result would no longer be recognizable as a
natural language -- you'd end up with would be effectively indistinguishable
from KIF. This was supposed to entail either that (i) the formal SUO is
*already* what you want, or (ii) what you want is impossible. But you seem
to be embracing the idea.
GH2> Suffice to say here that I see a third option, (iii) that
careful selection of rules for syntax and semantics can provide readable but
precise English, which would satisfy the criteria for a 1st order language.
> > > The SUO is currently instead heading down the road of becoming a
> > > highly symbolic ontology that will be very foreign and unreadable
> > to the average lay person.
> >
> > Like most any engineering model with the rigor needed to perform its
> > intended function.
>
> GH> As an engineer, one who has been an innovator in that role,
> and one who reads engineering history books as part of a hobby, I
> can't say I agree with you here. Some of those books are very clear
> and succinct in plain English, and could be used as design manuals.
I was thinking more in terms of technical mathematical models that are not
amenable to such translation. I think an SUO is akin to those -- though of
course, like any technical model, lots of helpful discursive annotation is
possible. But it's no substitute for the model itself.
GH2> Actually, some of the texts cover the mathematics, in
statements of fundamental principles precise enough for one to write that
relevant equations and calculate real results.
GH2> Certainly, I would accept that highly complex equations are
going to have their place, as they do in text books. Even then I find the
assumptions, and explications of the constituent portions, often come out
simpler for the reader if described in plain text, or possibly as a
dot-point list of sentences.
> GH> Coming from the other extreme, I would say it's possible to
> precisely and unambiguously explain in words every symbol and symbolic
> expression in English words. I'm not necessarily saying this would be
easy,
> mind you, but that's why I would prefer the adaptation mentioned above to
> avoid existing ambiguities and imprecisions.
Again, my claim is that you'd end up with, basically, KIF with perhaps a
superficially friendlier face. Consider:
(forall (?x)
(=> (Boy ?x)
(exists (?y)
(and (Girl ?y)
(Kissed ?x ?y)))))
This translates directly into
"Every boy kissed a girl."
But that won't do, right? Did each boy kiss some girl or other, or was
there some one girl that each kissed? The English is ambiguous where the
KIF statement is not. So we need:
"For each boy there is some girl such that he kissed her."
Things are already starting to sound pretty KIF-y and not very natural
language-y.
GH2> What about: "For every boy there is a girl he has kissed."?
(I've elaborated more further down)
But matters don't end there. "he" and "she" work in this case because there
are only two objects involved. But suppose we had instead
(forall (?x)
(=> (Boy ?x)
(exists (?y ?z)
(and (Girl ?y)
(Girl ?z)
(/= ?y ?z)
(Kissed ?x ?y)
(Discussed ?z ?x ?y)))))
What do we do with this? How about:
"For each boy there are two girls such that he kissed one of them and the
other discussed him with her."
Not exactly a lucid rendering. What's needed in order to clarify, of
course, is some sort of quantificational apparatus with variable binding to
make clear the meanings of the pronouns in the context, something like:
"For each boy_1 there is a girl_2 and another distinct girl_3 such
that he_1 kissed her_2 and she_3 discussed him_1 with her_2."
And things look KIF-ier still.
GH2> What about: "For every boy there is a girl he has kissed and
another girl who has discussed him with her."?
GH2> Actually, from the KIF text, it seems to me you mean kiss
and discuss as mutual activities, so what about: "For every boy there is a
girl with whom he has kissed and another girl with whom she has discussed
him."?.
Is this the sort of thing you have in mind? If not, you can't express in
your NL adaptation what can be expressed in KIF. IF so, I'll take KIF.
Note we can pretty up even the above KIF using sorted quantifiers, as John
Sowa regularly suggests:
(forall (?x : (Boy ?x))
(exists (?y ?z : (Girl ?x) (Girl ?y))
(and (/= ?y ?z)
(Kissed ?x ?y)
(Discussed ?z ?x ?y))))
I doubt you can improve much on this.
GH2> Actually, I think the suggestions I gave above aren't all
that bad, given the nature of what is being expressed. Basically it just
took a bit of time to get into a suitable mindset to translate these
expressions into more readable English.
GH2> At this point I am beginning to think it would be good to
have some real example uses for the KIF, to better gauge the difficulties
likely to be faced at the coal face rather than in the laboratory.
> GH Not a bit of it. I am proposing an absolutely precise
> adaptation of English. This would be largely a subset of the current
> language we all speak, for example each word would be restricted to
> only one meaning (preferably the most commonly used of its current
> meanings), but also:
> * where some words be allocated a modified
> meaning (precise, and preferably closely aligned to its current most
> commonly used meaning) where the current meaning is hazy ("manage" and
> "administer" may be examples where this would be appropriate, the
> former possibly being defined as the controlling through development
> of policy, and the latter as controlling through the application of
> policy); and
> * there would be some new words fabricated to cover
> useful concepts where these are either not currently covered by a
> single word, or, more commonly, where the current word used is a
> metaphorical application of an existing word with another main
> meaning(eg. "button", "mouse", "head", "monitor", etc.); while
> * there may also be means of indicating different uses of words, such
> as where the same word can act in more than part of speech (unless this
> can be reliably indicated by the specified syntax / grammar).
I think most of your proposal addresses possible ambiguities in particular
words. That is, granted, an important source of amibuity which, to some
degree perhaps, could be addressedin your NL fashion. But I think the real
challenge is in addressing ambiguity and other issues of logical form like
those above. Those are the phenomena that I believe will drive you toward
something equivalent to KIF.
GH2> I have already addressed the nature of at least some other
ambiguities previously, and posed the question as to whether, for example,
Russel's paradox isn't an inherent corollary of logic, much the same as the
fact that there are only analytic solutions for polynomials up to order
four. If so, then we need other ways of identifying and coping with such
inconsistencies, since they won't be able to be designed out of the SUO. I
can resend some correspondence on that if you want.
> > But that does not undermine the need for a formal SUO for computational
> > applications.
>
> GH> Of course not. What I'm saying is that, through such means as I've
> indicated above, the formal SUO can be based on natural language, and
that,
> if that is done (and done well), then it would achieve far wider
acceptance,
> and would provide have greater value to human kind.
Ok, I just didn't see anything about the validity, relevance, or importance
of the formal direction in your original post.
GH2> Chris, you may recall it was I who started raising concerns
about ambiguities in the first place, in the Scope & Purpose. I thought some
of that was in dialogue with you, and that it was you who put some of my
concerns to rest. I have discerned a great deal of trouble arises from
ambiguities and errors, and that's why I'm extremely strong on the need for
precision and accuracy. That's one of the reasons I see so much potential
benefit from the proposal I have been putting.
GH2> As I said earlier today, as a matter of relevance, the point
was also made on BBC radio last night, that one can already exploit the sort
of approach I am suggesting, when one wishes to use the translation software
available on the Internet. The observation was made, very much in line with
my proposal, that many people learn to use an unambiguous subset of English
in order to get their messages clearly translated into other languages. It
would be a relatively small step from there to achieving an ontology that is
rigorous, easily human and machine interpretable, and machine executable.
> > > However, I feel this means it will remain buried in the secret files
of
> > > industry designers and software developers, and probably not even of
> > managers.
> >
> > No secrets here; it will be a public standard.
>
> GH> Yes, technically you are quite correct, and this applies to
> many standards already. However, almost all of them may as well be
> industrial secrets, for all the majority of the populace are aware,
> even in "1st World" countries.
I suppose the SUO will be as big an industrial secret as, say, Newton's laws
or Maxwell's equations. I'm not sure "industrial secret" is the right
rubric here.
GH2> Actually, I would suggest worse than those examples. What
about the principles set out in the ISO 9000 series of quality management?
GH2> The difference is that, I am suggesting a user-friendly SUO
of the type I am proposing would be far more useful in everyday life than
Newton's laws or Maxwell's equations. For example, I would see it
facilitating people better understanding their legal rights, such as in
consumer disputes. To me, such benefits would be very worthwhile to society
these days. (######)
> > And far more accessible than many mathematically oriented standards.
> > (My better freshman students can translate proficiently into
> > first-order logic after 3 months of biweekly classes.)
>
> GH> Agreed, but that doesn't mean one closely based on natural language
> wouldn't be far user-friendlier.
If one were possible. My belief is that such a thing isn't, unless you call
constructions like the above "closely based on natural language".
GH2> As I said above, I suggest my alternative wordings are not
quite so bad. Let's see some actual application circumstances to flesh this
aspect of the discussion out more.
> I am proposing an SUO that would be far more widely accessible than
> one intended for society's intellectual elites.
I would caution against trotting out loaded terms like "elite" in this
context, as it suggests an intentional hoarding of resources for personal
power and gain; it would be profoundly unfair to tar anyone with that brush
without a great deal of evidence.
GH2> You did say "better freshman students". Also, tertiary
education is still only taken up by a minority as far as I'm aware. I also
meant élite from a point of view of being involved in making technical /
industrial decisions that are imposed on others unless, say, they have a
choice and use it to buy someone else's product.
It would be great if quantum mechanics and the calculus were more widely
accessible as well. But it just might be a fact that, like these areas of
science and mathematics, the content of the SUO will not be fully accessible
without a certain level of education. If so, the charge of elitism does not
stick; there is no *intention* to construct an SUO only for those who have
the requisite background education, it's rather that there is simply no
*option*. The responsible social tack in light of this is not to dumb down
the knowledge in question (which I believe is all an adaptation will be if
it is not equivalent to KIF), but to increase accessibility to the
educational prerequisites needed for understanding it.
GH2> I think my earlier words, which I have now flagged with:
"(######)" already address this point. Basically, again as I've said before,
I believe the 10 to the power of 21 neurones each of us has on average
represents an enormous amount of potential for ordinary people to grasp the
principles, so long as we make them readily accessible and interpretable.
There are many examples of one-time highly academic principles previously
the reserve of the cognocente that are now in the everyday arena. I believe
society by this increasing level of general levels of knowledge, and I see
easily readable presentations as good ways of assisting this. On the other
side, such presentations mean that we don't have to remember so much, and
can just keep the basic principles to underpin our thoughts, only calling on
the detail when particularly required.
> > > The main benefit I see from a "natural language based SUO" would be a
> > > greater clarity of logical implications of situations and devices
(both
> > > physical and conceptual) to the average person across societies and
> > > populations around the globe. I would include in conceptual devices
those
> > > that are educational, legal and sociological, so I suggest the
benefits
> > > would extend right across society.
> >
> > I wouldn't want to discount any of these. But again, the project you
> > envision is entirely orthogonal to the development of a formal SUO.
>
> GH> I don't agree, but please explain how what I am proposing
> differs from the development of a formal SUO. I am genuinely proposing
> an absolutely precise structure in both semantics and syntax, but one
> that is easily human and machine interpretable, and machine
> executable. What does the project involve that cannot be covered by
> such a structure?
Then I have no idea what you are proposing. KIF is a language with an
absolutely precise syntax and semantics. If you want a language just as
precise, it will not be any more readable to humans, at least not to any
appreciable degree, since what humans have the most trouble with are exactly
the things that a precise syntax and semantics *makes* precise (e.g.,
variable binding).
GH2> Let's flesh out the implications of my alternatives to your
examples near the top, before we come fully to grips with this issue. I
remain far more optimistic than your words here would seem to allow. I have
considerable faith in the benefits in the millennia of development
underpinning human language. I admit that, there was a large step backwards
in more recent decades, in the teaching of formal grammars, which has in my
opinion considerably disadvantaged younger people in analysing the structure
of language, and its logical potentials.
GH2> It's just a personal view, tho' I'm not the only one to hold
it, but I believe this fact has significantly disadvantaged society in
subsequent decades compared to what would be the case had those people been
afforded the education they lack in this area. In particular, I believe it
has robbed many of potential ability to accurately analyse situations and
develop sound responses to them. I could go on, but don't think this is
necessarily the best forum for what I believe is quite a significant matter.
Let me say, that, were it not the case, I believe there would be, among
other things, a greater demand for education in your field.
> A small example of the benefits I could see from a "natural language
> based SUO" could be a release from the frustrations of the current
> unintuitive and inconsistent behaviour of Windows, which many of us
> have to tolerate every day.
http://www.linux.org. ;-)
GH> Exactly, but how much market penetration does it have so far?
Well, over 1/4 of the server market, at least, and a growing share of the
desktop. But I get your point.
> GH> I am also reminded of various Unix purists I have come across in
> previous years, who felt that was the way of the future for the ideal
> operating system.
No, the way of the future is simply a *good* operating system that will
satisfy certain desiderata.
GH2> Agreed. However, I feel market exploitation has been used to
stretch out the development chain so as to maximise the need to update to
better software implementations. This feeling is in line with the story I
was told when young, that Gillette had a whole range of future product lines
stretching out into the future, so that the marketplace value could be
maximised through people having to change to new products. Comparing razors
over recent decades would seem to verify that story.
GH2> As a somewhat logical person who was a one time happy
WordPerfect user, I feel I am well aware of the extent to which logical
structures can facilitate use, such as the intuitive nested menu trees that
product provided, which minimised the amount of options requiring to be
renumbered. In recent times, I actually observe recent Microsoft releases
having more obtuse implementations, and suspect these are aimed at
discouraging keyboard use to access functions. For example, this version of
Internet Explorer has the command "<alt> V(iew), (G)o, B(ack)" to go to the
previous screen, while the previous version used "<alt> G(o), B(ack)", as
does this version of Outlook.
The desiderata for ontology require a formal, first-order langauge. Find a
prettier face if you can, but I believe its impact will be negligible. It
will either lack sufficient expressiveness, or it will be no easier to
comprehend.
GH2> Let's take the uses further, as discussed above, in order to
better assess this issue.
Highest regards,
-chris
--
Christopher Menzel # web: philebus.tamu.edu/~cmenzel
Philosophy, Texas A&M University # net: chris.menzel@tamu.edu
College Station, TX 77843-4237 # vox: (979) 845-8764