RE: SUO: Automated or Semiautomated Ontology Development
> -----Original Message-----
> From: owner-standard-upper-ontology@majordomo.ieee.org
> [mailto:owner-standard-upper-ontology@majordomo.ieee.org]On Behalf Of
> John F. Sowa
> Sent: Tuesday, 12 March 2002 2:46
> To: Bill Andersen
> Cc: cg@cs.uah.edu; SUO
> Subject: Re: SUO: Automated or Semiautomated Ontology Development
>
<snip>
> BA> Step up to the plate and write the code, and we'll see if it works.
>
> The two talks I mentioned discussed a lot of code that has already been
> written that does many of the necessary tasks:
>
> http://www.jfsowa.com/talks/negotiat.htm
>
> http://www.jfsowa.com/pubs/autotalk.htm
>
from the above links:
"Consider the verb support in the following sentences:
Tom supported the tomato plant with a stick.
Tom supported his daughter with $10,000 per year.
Tom supported his father with a decisive argument.
Tom supported his partner with a bid of 3 spades.
These sentences all use the verb support in the same syntactic pattern:
A person supported NP1 with NP2."
Neurcognitive processes are biased to the GENERAL notions of WHAT and WHERE
(object/relationship). These are particularisable to:
WHAT - WHO - WHICH
WHERE - WHEN - HOW
(WHY more often deals with values and as such comes 'later' in development
although it gets confused at times with being synonymous with HOW)
Thus all of the sentences above 'fit' at the GENERAL level:
[WHAT|WHO|WHICH] TOM
[WHEN|WHERE|HOW] SUPPORTED
[WHAT|WHO|WHICH]{tomato}{daughter}{father}{partner}
[WHEN|WHERE|HOW]WITH
[WHAT|WHO|WHICH]{a stick}{10000}{decisive arguement}{3spades}
Note the interdigitation of objects|relationships|objects|relationships etc
This reflects the interdigitations across the brain of left/right fields etc
and especially the interdigitations in the frontal lobes (and so
left|right|left|right connections of neurons from different hemispheres via
the corpus callosum etc etc) also note the ability for reversals (the verb
'to TOM' can be derived easily by conversion of WHAT to WHERE etc) as can
tomato - tomatoed, daughter - daughtered, father - fathered, partner -
partnered, stick - sticked....IOW CONTEXT allows for the elements of A/NOT-A
to swap)
Non-transitives are in the form of [WHAT|WHO|WHICH][WHERE|WHEN|HOW] as in
"Tom slept" and to these basics you add qualifiers ("Ugly Tom slept
peacfully") (that said you could focus on the deeds and say
[WHAT|WHO|WHICH][WHAT|WHO|WHICH] for "Tom slept" where the expression
'slept' is treated as expression of an object, a 'thing'...but more so there
is an implicit 'did' present:
[WHAT|WHO|WHICH]{did}[WHERE|WHEN|HOW] etc. such that we have in the above
TOM DID SUPPORT...or TOM DID SLEEP...)
The DICHOTOMY of what/where applied recursively allows for whats to look
like wheres and wheres to look like whats - a continuum of qualities emerges
from which to choose a category.
Local nuances in language can shift order - thus in German the
[WHERE|WHEN|HOW] are grouped at the end of sentences reflecting a grosser
chunking size of object|relationship and the relationships need distribution
to get 'meaning'. In French/Italian there is a play on adjective orders (the
qualifiers). In Japanese also the 'heads' change and there is an emphasis on
more implicit terms e.g. 'daughter' implies *his* daughter (Toms) unless
stated otherwise etc All of these languages have developed from trial and
error. We dont need these nuances for AI to AI communications.
Overall the focus is on objects|relationships, WHAT and WHERE, and the
application of that dichotomy recursively to give the GENERAL notions of
WHAT mapped to objects as WHOLES or PARTS and WHERE mapped to relationships
as STATIC or DYNAMIC. ALL LANGUAGE HAS DEVELOPED FROM THESE PRIMITIVES and
these primitives can be shown to be sourced in the neurocognitive processes.
Understanding the manner in which the mindless brain processes
objects|relationships indicates the general level of information processing
upon which all else is founded. IOW you could develop a language for AI
systems that has nothing to do with English or Latin or Greek or Chinese etc
etc etc as long as it deals with objects|relationships; why copy the brain?
why 'localise' AI language? just copy the methodologies - the brain has
spent x million years 'mindlessly' developing - we can do it quicker since
we are, to some degree, mindful.
In our own species, if allowed, each generation would create its own
language - so to then AI systems. Feed them the basics and let it 'evolve'.
regardless of the manner of expression they will all talk 'objects and
relationships' where and object is anything I can use as 'WHAT' and a
relationship anything I can use as 'WHERE'. CONTEXT will determine what
becomes preferred as a 'what' and preferred as a 'where'; the general
what:where template allows for both (and even oscillations where we deal
with paradox.)
The neurocognitive research means we need to review such concepts as
transformational grammar etc
Chris.
------------------
Chris Lofting
websites:
http://pages.prodigy.net/lofting
http://www.ozemail.com.au/~ddiamond
http://www.eisa.net.au/~lofting
Lists:
http://www.yahoogroups.com/group/semiosis
http://www.yahoogroups.com/group/ichingplus