DM: A Program For Grade One Reading

I would appreciate any comments on the exchange with Professor Minsky,
below. We associate our verbalizations (eg in Q-A sessions re reading
comprehension) very much with consciousness or sentience. Could it be that
we are no more cognizant of the calculus for word-usage than a person
talking in his sleep? With millions of experts on the planet who deal with
one aspect or another of grade one reading, can we expect to find a set of
known rules for giving correct answers to questions which could be
programmed into a computer?
FWP

---------------------------------------------------------------------------

Date: Sun, 17 Sep 2000 11:45:05 -0700 (PDT)
From: Franklin Wayne Poley <fwpoley@vcn.bc.ca>
Reply-To: EDTV-Robotics-State-Of-The-Art@egroups.com
To: Marvin Minsky <minsky@media.mit.edu>
Subject: [EDTV-Robotics-State-Of-The-Art] Re: the "Human Equivalency"
     Question (Which is what the proposed ed tv program will be built around).

On Thu, 14 Sep 2000, Marvin Minsky wrote:

 > At 12:38 PM -0700 9/14/00, Franklin Wayne Poley wrote:
 >
 > >On Tue, 5 Sep 2000, Marvin Minsky wrote:
 >
 > > In cases where humanoids are still inferior, does it 'only' take
 >
 > >a greater deployment of money and labour to close the gap or do we
 > need to
 >
 > >INVENT something new?
 >
 > Yes, money by itself won't help, until we have better ideas about what
 > to spend it on.

clip

OK, let's forget factor analysis for now. Let's take all the tough
problems in AI and set them out and then analyze them by a different
method...let's call it common sense analysis since you invoked common
sense in an earlier email. Here is one you gave in your email of
Aug. 29/00:

"We have no programs, to this day, that can answer simple questions about
what they've 'read' in a first grade story book."

I am on a datamining list which is populated mainly by computing
science-statistics experts (as best I can tell). From the discussion on
definitions I think this qualifies as "data mining from text". I will put
this problem out to them too. In common sense terms how would I go about
solving it?
    First I would seek out the experts on teaching grade one
reading...specialists in early childhood education. As smart as they might
be I don't think linguists and psychologists in general can help much when
it comes to working out the specifics of this problem.
    My question would be whether they know the rules which are used when a
child learns to answer questions about a text at that level. Maybe they
don't, as astounding as that may sound. Maybe even with millions of such
educators on the planet we (humankind) still don't know how a
six-year-old reads a book. What rules does the six-year-old follow to
answer the questions which are asked?
    If that is so, much of education at this level is going on at an
unconscious level. If it were conscious I expect we could verbalize it.
Now here is where philosopher-psychologists with clinical backgrounds like
myself could be helpful. We could have a try at "raising
consciousness" for those grade one teachers (and their students) and we
could do it online even. So we would go from unconscious teaching-learning
to conscious teaching-learning to verbalizing of the teaching-learning to
writing a computer program for it. When Skinner said, "If it can be
verbalized, it can be programmed" he probably didn't have computer
programs in mind. But I think he meant programming in general and
"Skinner's Dictum" seems to apply pretty well to computer programming too.
If we can verbalize the rules for answering questions in the grade one
reader I expect those rules can be programmed into a computer.

clip

• Follow-Ups:
  • DM: Re: [ML] A Program For Grade One Reading
    • From: Ron Blue

• Prev by Date: DM: ICML-2001...Interdisciplinary Machine Learning Research (fwd)
• Next by Date: DM: Re: [ML] A Program For Grade One Reading
• Prev by thread: DM: RE: Where to find references on Data Mining process
• Next by thread: DM: Re: [ML] A Program For Grade One Reading
• Index(es):
  • Date
  • Thread