linguistic & mathematics

Vaidix Vaidix at AOL.COM
Mon Jan 19 11:43:41 UTC 1998


Dear Mr Thillaud,

>(I'm not convinced the fuzzy set theory is a powerful tool in
>linguistic, I believe that's just a attempt to give some obscurity (hence
>seriousness) to simple facts.)

Not yet, it is a new field in infancy.

The digital computers are based on a design that uses some specific states
"on" and "off".  Everything else in the digital computer is definite.

The brain and other neuro material in the body are made up of millions of
neurons.  These neurons do not work like digital switches, but have their own
method of computing (called neuro-fuzzy computing).  The neurons do not work
as fast as the digital switches of a computer, it is like comparing a tortoise
to a space rocket, but the neuro computing (of even a two year old child) can
achieve miracles like pattern recognition (such as recognizing the caricature
of a cow as a cow) which even digital computer can not do to this date.  What
is the secret of this miraculously great computing power of the brain?

There are two aspects to this.  Neuro and Fuzzy.

Let us talk of "Fuzzy" to begin with. Each neuron knows pretty well, the
futility of trying to digitize an observation and accurately measure it, there
is just not enough time to do it, considering the urgency of the situation the
organism faces.  As a survival instinct, the neuron assess each event and
categorizes the event into one of the "few" known "linguistic" states of mind.
Let us say there is a neuron located in a certain place in the brain (which
can be verified by scientific experiments) which is an expert in assessing
distances.  This neuron never attempts to measure the distance of a speeding
vehicle in exact meters and milli meters to begin with.  It just puts the
distance into some knows states such as "very near", "near", "a little far" or
'far" etc.  The brain then uses these fuzzified inputs in its processing.  If
the results are not satisfactory, next time the neuron learns to improve its
assessment of distance.

To actually measure the distance in meters, we have to probably go to other
neurons in a different place in the brain, which are experts in that field.
Depending on context, these neurons may translate "very near" as ten meters,
"very far" as 100 meters etc.

There would be other neurons that decide whether there is a need to measure
the distance more accurately or whether to stop measuring and take necessary
action considering the urgency.  These neurons classify the time available
into linguistic states such as "ample time available", "some what urgent" or
"very urgent" etc.  The list is endless.  It is all a team work by neurons
which are experts in different aspects like this.

The other aspect of the miracle of the mind is "Neuro".  The word neuro
denotes a network of neurons.  A single neuron can not achieve much other than
express its own feelings in a chemical or electric language.  A network can do
wonders.  "Making connections" between neurons has the linguistic effect of
relating different concepts.  "Transmission" of signals is the linguistic
equivalent of deduction.  Readjusting the connections between neurons
depending on success or failure of present action is the lingustic equivalent
of "Learning".  The list is endless.

Combining neuro and fuzzy concepts, computer scientists are able to write
programs that can do near-human things like pattern recognition (such as
signatures).  These are the things that can not be achieved by tradition
computer programming.

When computer scientists can benefit from linguistic experience, why can't
lingusts benefit from neuro-fuzzy theories?

If a neuron assesses each event into certain lingustic states, conversely each
syllable used in linguistics in any language is the result of such assessment
by neurons. That is the connection.  The field is brand new, and a lot more
research is needed before it can be put to practical use, but we can make a
beginning.

Bhadraiah Mallampalli
http://www.fortunecity.com/roswell/bailey/64 (Vaidix).





More information about the INDOLOGY mailing list