[From: Oded Maler (931027)]
[Bruce Nevin (Wed 931027 10:08:12 EDT)]
(A lot of nice stuff deleted)
* I think this is at the bottom of an apparent paradox. The paradox
* is this: HPCT promises to account in detail for all human behavior (and
* for much more). Students and investigators of HPCT can account for their
* inability to gain wider acceptance, understanding, and application of the
* theory, and the account is insightful. However, the account is not
* useful for bringing about the avowed goal of gaining wider acceptance,
* etc., for the theory. There is much complaining about the discrepancy
* between this goal and actual perceptions, so there appears to be chronic
* error in a lot of control systems participating in this net. Perhaps the
* bickering between Greg and you, Rick, is due to your respective
* reorganization processes having arrived at different ways of reducing the
* error--Greg's to build bridges, yours to ignore it, maybe it'll go away,
* and hey it doesn't really hurt--I mean, what error, really!
This practical failure in the small microcosmos of academia, should
not, of course, put some doubts in the correctness of the slogan
"more PCT knowledge ==> better real world". (I'm just teasing).
A-propos phonemes and all that. Two years ago, following some discussions
on CSG-net I had an idea of a useful tool to improve pronounciation
(a great problem for everyone living in France). The idea is very
simple, the user is asked to say "en" (not to confuse with "in" "on",
etc.), a basic voice processing system extracts features from his
performance and compares it to some templates of correct pronounciations.
The distance (in the vocal space) between his articulation and the
correct one is displayed graphically, and the goal of the user is
to bring his point to the correct target. Since people have a lot of
expertise in visual feed-back, I believe that given such a device
they could find the best direction in the articulation space, in order
to reach the target. (Maybe I'm naive).
I had some talks with speech people in my previous lab about details,
e.g., transform the sound into 7 dimensional Fourier spectrum,
segmentation problems, etc., but like many ideas it was dropped down,
so I present it here to anyone interested.