Underlying Formalisms LO6288

John Zavacki (jzavacki@epix.net)
Fri, 29 Mar 1996 07:51:50 -0500

Replying to LO6265 --

John O'Neill wrote:

>Replying to LO6252 --
>- can we socially construct knowledge (language) based on an underlying
>formal language representation?
>My understanding of the work in natural language processing in AI
>is that
>they wanted to start by pre-defining all possible relationships within
>language. For example, how many possible ways are there for constructing
>a sentence?

There have been many approaches to natural language processing in AI. As
someone who did research in this field at a time when computers were not
readily available, I came to a few conclusions, one of which dealt with
the "possible ways for constructing a sentence":

Grammar is finite. There are a finite number of relationships across the
syntactic elements. Meaning is not. The personal learning history varies
tremendously, attaching emotion, partial memory, and illusion, allusion,
and delusion to the same sentence based on social context, emotional
state, etc. Discussable, but probably not programmable.

>Even if this is possible, one must question what value is imparted
>to a system that uses this approach. An alternative is to construct the
>relationships as they are defined and used by a social group (after
>all,language is about aiding communication between sentient beings). In
>this way, the relationships in a language are constructed as we go.

Relationships (linguistic or otherwise) are dynamic and we do "construct"
(or at least interpret) them as we go. The problems with universal
definitions of these relationships remind me of (paraphrase coming) Ortega
y Gasset's definition of the task of philosophy:

"....to describe the world."

John Zavacki
The Wolff Group
900 James Ave., Scranton, PA 18510
phone: 717-3416-1218  fax: 717-346-1388

Learning-org -- An Internet Dialog on Learning Organizations For info: <rkarash@karash.com> -or- <http://world.std.com/~lo/>