RE: [topicmaps-comment] Everchanging subjects [ Notionshaveexistence...]

[Paul Stephen Prueitt <bcngroup@erols.com>]

Yes, I do follow Hofstader work, and based a lot of my work on text element
similarity analysis that I tried to get built into automated
declassification technology in 1995 - 1998.

I will say, however that the notion of large maps does not feel right to me.
If we have a computationally emergent process that produces a small and
situationed graph, then the graph might have less than one hundred nodes
(topics) and perhaps even less than 10.  This would correspond to the moment
by moment presence of concepts in a human mind.. not millions or billons or
even one for each word in a dictionary.

In my mind I see the emergent small topic map as something that is created
via a second order control system and a human awareness - but not as a
retrieval of a small part of a much large topic map - like system such as
Cyc.

So, whereas I agree that similarity analysis like Hofstader and his few
students (at NMSU mostly) is very important to the specification of semantic
linkage, this excellent work in no ways makes any progress towards the
problem of situatedness.

David Dodd's notion of contextual situatedness may be close to mine - I can
not yet tell.  He comes from a technical background where things seems to be
said mostly in XML type jargon.  My background is pure mathematics, physics
and neuropsychology.

I have discussed recently the notion that situatedness must be properly
related to the physics of space and time and that the simulated models for
(for example Complex Adaptive Systems) are not "stratified" in the sense
that the group developing stratified complexity theory is defining.

These thoughts expressed above are formative and still under development.



-----Original Message-----
From: Thomas B. Passin [mailto:tpassin@home.com]
Sent: Friday, October 19, 2001 5:31 PM
To: Topicmaps-Comment
Subject: Re: [topicmaps-comment] Everchanging subjects [ Notions
haveexistence...]


[Paul Stephen Prueitt]

> But my attempts to create software is not important.  What is important is
> that a stratified complexity paradigm be developed and adopted that has a
> route to computational emergance.  This route, in my opinion can not be
> completely made within a data structure using solely computational
> processes.
>

Paul, do you see the work of Doug Hofstader and his group as having a close
relationship to the things you are talking about?  I'm referring to the
combinatorial discovery programs like the series solver and so on, that
build temporary recursive structures drawn from a fixed store, using scoring
systems and probabilistic methods to construct additional structures?

On another note, I was just thinking about what might be the complexity of
real world maps - I mean ones in peoples' heads, not topic maps per se - and
how could be discover this?  Here's one line of thought.  Apparently
American high schools students know some 60,000-80,000 words, maybe many
more if they are good readers (according to Paul Bloon in "How Children
Learn the Meanings of Words").  Suppose we were to make a topic map of this
vocabulary.  Then there would be at least 60,000-80,000 topics right there.
But people have many associations for each word, and in fact different ones
for a word used in different contexts.  Say on the average a person has 10
associations for each word (it would be many more for familiar words).  Now
we are close to a million things in the map.

No doubt this is just scratching the surface, and already we are up to a
gigantic map.  Pretty impressive!

Cheers,

Tom P


----------------------------------------------------------------
To subscribe or unsubscribe from this elist use the subscription
manager: <http://lists.oasis-open.org/ob/adm.pl>