[202] in bcs-newton

home help back first fref pref prev next nref lref last post

Newton Handwriting Recognition

daemon@ATHENA.MIT.EDU (Eric Celeste)
Tue Aug 24 15:12:35 1993

Date: Tue, 24 Aug 93 15:13:24 -0400
From: Eric Celeste <efc@wonder.mit.edu>
To: ron@pedi.ama.ttu.edu (Ron Jones)
Cc: bcs-newton@world.std.com, white@nest.enet.dec.com
Reply-To: efc@MIT.EDU

> Can someone tell me if the Newton MessagePad deciphers only PRINTED 

> LETTERS or does it decipher CURSIVE handwriting.

Newton can translate both printed letters and cursive handwriting. The  
Newton recognition software does not just look at the image of the character  
you've written on its pad... it also pays attention to the order and timing  
of the strokes of your handwriting. In my informal observation of friends  
trying out the Newton, I've actually found it to be _more_ reliable at  
translating cursive writing than print. My personal theory on this is that  
we tend to learn cursive a little more formally, and the very fact that you  
write all the letters in a continuous flow imposes a more predictable order  
and timing on the strokes.

> I read somewhere that Newton interprets more word-by-word rather than 

> letter-by-letter...

Newton's recognition software can be forced to turn any word you write on  
the pad into a word from its word list (a kind of dictionary), or it can be  
set to turn words into words that are not neccessarily in its word list. The  
latter would be more like "character" recognition, though Newton still tries  
pretty hard to generate words it knows about. So, actually, Newton's  
recognition software can interpret either way.

...Eric

home help back first fref pref prev next nref lref last post