Re: [dynamic_gui] notions of covers

[Kurt Cagle <kurt@kurtcagle.net>]

Paul,

> Where is the first meeting at..  I know Daniel Rivers-Moore and have
talked
> with both Steven Newcomb and Michel Biesunski, starting after a
presentation
> I made at Knowledge Technologies 2001 on Knowledge Management and
Knowledge
> Science.
I've occasionally corresponded with both Michel and Steven, though I would
say that if they were hard-pressed they probably don't remember me -- they
are many light years ahead of where I am wrt to knowledge representation
theory, so most of the interactions I've had with them have been queries.
I've heard of Daniel Rivers-Moore, but I'm not well familiar with his work.

> Kurt, I was disappointed in XTM as the community seemed to not really
> understand why it is important to have a formal hook for non-addressable
> subjects.  Doing ontologies is not sufficient, and these are not living
> systems, as natural langauges are.

I've been lurking on Michel's Topic Map lists for nearly a year, though I
will admit that I spend far less time following it now than I initially did.
My sense with XTM has been similar to yours, I suspect, and admittedly is
one of the facets that's bothered me as a developer as opposed to a
theoretician. Topic maps struck me initially as attempting to get a decent
handle on the whole issue of ontology, but after a while it seemed to me
that XTM had swirled off into irrelevancy, which makes me seriously wonder
whether ontology itself is at a layer of abstraction that makes the process
of understanding the basis of knowledge too complex. We build grammars and
vocabularies, yet in so creating such structures we (as you so aptly point
out) fail to take into account that language itself is dynamic and richly
textured. Perhaps the key to the next stage of KM is to recognize that idiom
is a fundamental part of any language, whether natural or more highly
structured. However, I'm not a linguist, nor do I even play one on TV.

> Kurt, are you going to Knowledge Technologies 2002
> (www.knowledgetechnologies.net)?  If so might you co-organize a session on
> the human dimension to knowledge representation?
I'm probably out of my league there, though with it being in Seattle I could
at least try to get something in (I live in Olympia, so if I can get past
the admission costs I'm home free). We're just past the deadline for
submission, unfortunately, so I'm going to have to come up with something
pretty damn compelling. Let me see what I can do.

> I am on the organizing committee.  But I only have until the end of this
> month to organize the KM dimension to the conference.
>
> I grin as I see you that you said:
>
> "I tend to suspect
> strongly any measure which goes to zero with an infinite number of items,
> and ultimately I think that density is simply a handy Euclidian metric
that
> smooths out the largely fractal nature of reality."
>
> and ask that you look at ::
>
> http://www.ontologystream.com/IRRTest/Evaluation/ARLReport.htm
Interesting. I'm definitely reviewing it.

> ...Yes it is true that Cantor and Godel both made mistakes...
To be sure. Consider this point, though. Godel's fundamental thesis was
that, paraphrased dramatically, no rules based engine could effectively
model all assertions within the context of those rules, and conversely that
there will always be assertions that cannot in effect be modelled by a rules
based engine. Yet to a certain extent isn't this exactly what XTM is
attempting to do?

I'm probably reading into this more and is there, though this also raises a
significant point about the role of analogy within systems. An analogy is a
structural similarity, not a semantic one. This point is important to me,
because of what I mentioned in my last missive -- if you build a graphical
interface based upon semantic relationships, then you become very heavily
dependent upon the hierarchy of those relationships. Think, for instance of
a typical UML diagram or class model. Such classes are semantically
interdependent, but are not necessarily (and probably aren't, for that
matter) structurally similar. There is no real ability to build analogy into
such a system, where analogy can in effect be defined as utilizing
structural similarities between two distinct schematic hierarchies even when
the hierarchies aren't semantically related. Analogous systems are usually
not truly isomorphic; at some point the analogy breaks down, because
semantic differences do come into play, but a significant amount of learning
takes place largely through the use of analogies to create a semantic map up
to a certain point, then a second semantic map that creates a second order
approximation of the system in a specific domain, and so forth.

What comes to mind is (analogies at play again) the ontologically equivalent
of Fourier Transforms. Most people do not absorb whole structures intact.
Instead, they create holographic approximations, generating a first order
representation based upon a broad set of metaphors, then honing those
representations through ever more constrained metaphors, while
simultaneously removing the semantic import of the analogy as information
becomes more highly refined. In essence, much of learning becomes then the
ability to throw away the semantics of the metaphor once the structure has
become incorporated into the holographic core.

I may be completely off base with this; it's not something that I've seen
formally discussed, though admittedly I'm still a novice in the KM field.
However, I think analogous holographic systems are very definitely the key
to building representations of information as well, getting back to the
dynamic GUI aspect of the discussion, because I feel intuitively that so
long as the media representation of an object in XML space is predicated
upon semantic terms rather than analogous or structural terms we'll be
struggling to build effective interfaces.

-- Kurt Cagle