The old GRAIL recognizer in the late 60s was pretty much 100% and 
buffered the recognition so you could put in characters as fast as 
you could draw them -- some pretty impressive speeds were achieved 
with this.
      And the Engelbart 5 finger keyboard+mouse way of navigating and 
inputing was quite similar: it was perfect and pretty fast.
      However, everyone who got fluent in these systems agreed that 
there should also be a keyboard for inputing of paragraphs, and there 
was pretty much universal agreement that the UI should be set up so 
that one tended to stay in one setup (regular keyboard) or the other 
(nav + character input) for as long as possible. This is why the 
Dynabook model always showed both a pen and a keyboard.
      The really important thing about the penbased systems of the 60s 
was that the ideas of inputing text and pushing buttons were quite 
secondary to the importance of being able to draw shapes and have 
them recognized and be able to hook them together graphically and 
semantically. Its this dimension that is sadly lacking in most 
penbased systems today (especially wrt PDAs, etc.).

The stuff that Nathanael and Ned are doing is quite impressive and is 
a wonderful throwback to the much more expressive systems of long ago.

Cheers,

Alan

At 1:36 PM -0500 6/13/03, Aaron J Reichow wrote:
>On Fri, 13 Jun 2003, Ned Konz wrote:
>
>>  Still, no matter how accurate your gesture recognition is, it's
>>  painful to enter any large amount of text when you're just doing
>>  per-stroke recognition. Most of us don't do clean, disconnected
>>  strokes when we write fast. We'll be experimenting with feeding the
>>  strokes captured by my Ink Morph to Genie in the background (perhaps
>>  using a dictionary like M$ does), but I don't expect it to work very
>>  well.
>
>One possibility I've always wondered about is using an external
>recognizer, feeding it the information from your Ink Morph, for example.
>(is this available anywhere?) On desktop Windows and CE, Squeak could feed
>it to CalliGrapher/Transcriber- it has an API for exactly this sort of
>thing.  I've never done it for Dynapad on WinCE/PocketPC, but it could be
>a way to provide real HWR to folks on those platforms.
>
>In this InkMorph+Genie thing, are you planning on attempting real HWR at
>all?  Trying to disconnect the letters of a word and then feed it to
>Genie?  While I've never achieved much of a speed with any sort of
>stroke/character recogntion, I can achieve 50 WPM using Newton OS 2.x HWR
>or CalliGrapher.
>
>Regards,
>Aaron
>
>--
>   "life, probably the biggest word i've ever said, that says a lot,
>   because there's a whole lot of words inside my head.." :: atmosphere


--