🚀 go-pugleaf

>Subject: Re:  Interactive videodisc project--American Sign Language
>
>Well, it was pretty interesting.  You could also advise your friend that
>there already exist computer-driven ASL output _and_ input systems!  Output
>means that a user types a word on the keyboard and a mechanical hand moves
>to form the ASL "characters", which a ASL-competent person "reads" with by
>holding it with _their_ hand.  Input means that the ASL-competent person
>wears a "data glove".  This conveys the position of the hand and fingers to
>a computer which is running a program that recognizes the ASL word or char-
>acter, then produces the corresponding ascii text and synthesized speech.
>
>I saw this on PBS about 6 months ago.
>
>
Actually, what you saw was a robot hand that fingerspells -- this is not
ASL. Fingerspelling is a manual representation of English graphemes (letters
of the alphabet) -- one handshape for A, one for B, etc. Of course, when
you string fingerspelled letters together, you get --- English. ASL is not 
English.

As a matter of fact, though, the device you saw has some parsing problems.
By that, I mean it has problems knowing where fingerspelled words begin
and end. To solve this problem, the user (the one wearing the DataGlove) adds
a special nonce handshape at the ends of words. One other problem -- the
device has problems reading letters in context. Just as speech sounds are
modified in context (which is why speech recognition is such a difficult
nut to crack), fingerspelled letters are contextually modified (the fancy
word for this os "coarticulation", and it happens in speech, fingerspelling,
and even in signing ASL).

Still, though, the device is very interesting. BTW, I have tried the DataGlove.
It's a very intriguing piece of equipment. Very lightweight, doesn't interfere
with range of motion, etc. We have considered (down the road!) trying to use
the DataGlove as an input device for the ASL Hypermedia Dictionary. If the
user wanted to look up an ASL word, she would just put on the DataGlove and
sign it!  Of course, I expect we would have just as many problems attempting
to implement this as folks in speech recognition have. So for now, we are
just putting our efforts into developing a useful dictionary.

Sherman Wilcox
Dept. of Linguistics
University of New Mexico