[Top] [All Lists]

Re: [ontolog-forum] Foundation ontology, CYC, and Mapping

To: "[ontolog-forum]" <ontolog-forum@xxxxxxxxxxxxxxxx>
From: "John F. Sowa" <sowa@xxxxxxxxxxx>
Date: Mon, 01 Feb 2010 20:38:20 -0500
Message-id: <4B67820C.8050007@xxxxxxxxxxx>
Ronald, Rob, Doug, and Pat,    (01)

Today's notes are getting at the heart of the matter.    (02)

RS> ("Philosophy" paper) "But by always translating one lot of signs
 > into another lot, we never bridge the gap between sign and reality.
 > We must discard the almost religious desire for purity, put our
 > heads above the wall separating the technical from the human aspects
 > of information systems, and embrace the untidiness of human beings
 > who alone can link signs to reality."    (03)

I sympathize with the conclusions, but I'd rather not prejudge the
issues by claiming that some ineffable, mystical human abilities
could never be simulated on a computer.  We need further analysis.    (04)

RS> I certainly did not intend disparage efforts to find formal
 > methods for handling semantics (why else would I have worked so
 > long on my own project?), I wanted to emphasise the limits of
 > that approach.    (05)

Analyzing the limits is crucial.  The concluding chapter of my 1984
book was "Limits of Conceptualization" and Chapter 6 of my 2000 book
was "Knowledge Soup".  In 2004, I discussed some of issues further:    (06)

    The Challenge of Knowledge Soup    (07)

This article analyzes the issues that have not been adequately
handled by current formalisms.  That challenge must be addressed.    (08)

RS> I prefer to work on concrete case materials to which we could
 > apply both approaches and test for points of compatibility and
 > incongruity.  This would quickly reveal whether our goals are
 > too far apart or not.    (09)

Case studies are an excellent means for testing ideas.  Could you
please post the details of "Japan Wines Inc" on your web site?    (010)

RF> ... why can't "background knowledge" too be represented in
 > patterns among text, patterns of word use etc?    (011)

It can be, but an enormous amount of detail would be lost.
For some applications, that detail might not be critical.
But for others, it may be the central focus of the subject.    (012)

Consider the examples I cited in my last note:    (013)

JFS> ... chess, hiking, programming, eating, cooking, chemistry,
 > surgery, driving a car, and football.  For most of those
 > subjects, a very large part of the background knowledge will
 > *not* be expressed in either verbal patterns or some version
 > of logic.    (014)

For chess, the details are expressed in chess notation, which is
a kind of specialized logic that could be translated to a subset
of FOL.  But the verbiage about chess is usually stated in forms
that are rarely, if ever, sufficiently precise to analyze a game
of chess.    (015)

DF> When reasoning about chess, game playing and spatial contexts
 > are needed as well as a basic temporal context, but no contexts
 > about physics or physical objects, chemistry, biology, or human
 > relations are needed. A special microtheory with the rules of
 > chess would be needed as well.    (016)

I agree.  The patterns in the chess notation are critical, and
not likely to be processed adequately by tools designed for NL
or by combinations of vague terms taken from Longman's list.
Programming is similar to chess, and the verbal patterns built
on Longman's word list are as clumsy for a precise analysis of
computer programs as they are for analyzing chess positions.    (017)

For hiking, both verbal commentary and FOL are far less effective
than an ordinary map.  But even the usual hiking maps aren't
sufficiently detailed for guiding a human (or a robot) over the
rocks, tree roots, muddy spots, loose gravel, etc.  The language
that a hiker would use or its translation to FOL would be nearly
useless, compared to full 3-D imagery with the option of zooming
in for detail or zooming out for a wider view.    (018)

DF> Football requires game theory, but also naive physics, naive
 > human anatomy, and sports contexts.  Cooking would use naive
 > physics and food ontologies, but not game theory or anatomy.    (019)

Football requires far more than naive physics.  Those qualitative
versions of physics were proposed for understanding the verbal
patterns in ordinary language.  But a robot that could play
professional football would need *real* physics at a precise,
detailed level.  A robot that used only "naive" physics might
flip a burger at McDonald's, but it wouldn't be an expert chef.    (020)

Surgery, driving, and football are similar to hiking in their
dependence on imagery and physical actions that are extremely
difficult to describe and reason about in words or logical
formulas.  Some mathematical techniques can handle them,
but they use continuous methods that are just as difficult
to describe in words as the original images.    (021)

Eating has aspects of imagery and physical action plus
the senses of smell and taste that are even more difficult
to describe than visual images.  Cooking and chemistry are
mixed fields that have some aspects that can be described in
logic and others that require imagery and continuous math.    (022)

PC> All logically compatible "ways of looking at the world" can
 > be accommodated in the FO, or if those views are not primitive
 > elements, then in some extension.    (023)

Note the discussions above.  The issues of dealing with imagery
and physical actions are handled by continuous math, not by FOL.
The smells and tastes cannot be adequately characterized by
any version of logic we have today.  Note the descriptions
of wines as "having notes of berries and fruit."  Those are
extremely vague statements that even a wine expert couldn't
interpret to the precision necessary to identify a wine.    (024)

PC> The principle that I seem not to have been able to adequately
 > convey to all is that the FO as proposed provides a set of
 > primitive (non-decomposable) ontology elements to describe more
 > complex ontology elements - and to translate among different
 > domain ontologies.    (025)

I don't believe that you ever seriously addressed the problems
discussed above for chess or computer programs, much less the
far more complex fields of hiking, surgery, driving, or football.    (026)

Suggestion: You might consider collaborating with Ronald in the
case study.  If you can show that your proposals for an FO would
be useful for that case study, it might help you state them
in a more persuasive form.    (027)

John    (028)

Message Archives: http://ontolog.cim3.net/forum/ontolog-forum/  
Config Subscr: http://ontolog.cim3.net/mailman/listinfo/ontolog-forum/  
Unsubscribe: mailto:ontolog-forum-leave@xxxxxxxxxxxxxxxx
Shared Files: http://ontolog.cim3.net/file/
Community Wiki: http://ontolog.cim3.net/wiki/ 
To join: http://ontolog.cim3.net/cgi-bin/wiki.pl?WikiHomePage#nid1J
To Post: mailto:ontolog-forum@xxxxxxxxxxxxxxxx    (029)

<Prev in Thread] Current Thread [Next in Thread>