Friday, February 27, 2004

Restoring Graffiti One type print recognition to the Graffiti Two (Jot) Palm devices

Google Groups: View Thread "Tungsten E and print recognition: what's your experi..."
I went through TealScript and ruthlessly deleted all the
variable strokes from my personal profile. The only strokes allowed
are now the ones I use -- the advanced high speed but often
undocumented G1 strokes documented in Pogue's O'Reilly books. I
defined all of my strokes as "standard".

I also deleted many of the "distorted" strokes.

Once I'd paired down the base strokes, I used TealScripts "train
profile" exercise from different angles to create a library of
distorted variants of my base strokes.

I think I could further refine this by tweaking the TealScript squelch
and cutoff settings. Shockingly the best advice I could find seems to
be in the manual.

Based on the manual I have set Squelch and Cutoff to 0 for all of my
"standard" strokes. If one of my standard strokes is being applied
when I don't want it, then

- if it's a simple stroke I'll boost squelch
- if the misapplied strokes is not simple I'll boost cutoff - esp. if
it's distort mismatch

For my distort entries I'll experiment with setting cutoff to 5-15 ...

One of the miracles of the US Robotics PalmPilot was that the pen character input worked. It worked because of Graffiti, a novel form of printing. The brilliance of Graffiti was that it asked the human to adjust to the limitations of the computer. With a bit of practice one could get pen input to work.

Unfortunately few were willing to learn Graffiti. Xerox won a patent fight about its origins, but it was a false victory. Palm abandoned Graffiti in favor of Jot -- a more familiar but singularly ineffective form of pen input.


No comments: