From: R. Dylan Stewart (Bob_Zimmerman_at_myrealbox.com)
Date: Mon Feb 16 2004 - 13:03:44 PST
> All,
>
> I've been curious about the TabletPC's handwriting recognition in relation
> to the Newts so I did some research, even posting a bit over on MS's
> TabletPC forum. The thing that leapt out at me is the fact that it's static
> - it has no facility for learning how your write! Here's a comment from one
> of the MVP's on the TabletPC forum:
>
>> The handwriting is not trainable. Period. There are hundreds of
> thousands
>> of samplings within the handwriting recognition, with many years of
>> development put into the project. It really does get better as the user
>> gets used to using their Tablet PC.
>> --
>> Chris H.
>> Microsoft Windows MVP/Tablet PC
>
> Notice the last line - it gets better BECAUSE THE USER GETS USED TO USING
> IT. IOW, just like PalmOS, you have to change how you write to improve your
> chances that your handwriting is recognized.
>
> In the grander scheme what strikes me is how we're still producing products
> that force the end user to conform to how the computer works, instead of the
> other way around. In many respects I mean this in a much broader sense, but
> AFA handwriting recognition goes that should have been one of the top 3
> development goals.
>
> I'm curious if anyone else has any thoughts on this or more knowledge about
> the TabletPC's handwriting recognition engine?
>
> Thanks.
>
> -Henry
Well, I tried a Tablet PC down at the local Fry's. I've also been using
the HWR engine included with Microsoft Office since Office xp. The HWR
is quite impressive, really and it's nothing like Graffiti at all. It
got my name (Dylan) on the first try. My Newton tends to recognize is
at pylon for some reason.
I don't like the lack of adaptive recognition. They have adaptive
speech recognition, so why not HWR? Anyway, I had to change my writing
style just a little bit to get my Newton to recognize my handwriting
better. These changes include writing my 'g's as a single stroke rather
than like a j with a c on top and writing my '4's as two strokes instead
of one. The Newton would consistantly recognize those two characters
incorrectly. The g would show up as jc and the 4 would be a 9. I'm
sure that if I gave it enough time, it would have learned how to read my
'g's and '4's, but it was easier to just change how I write.
I've noticed that since I got my Newt, my handwriting has become MUCH
more legible. I suppose that I've been forcing myself to write more
cleanly so as to get a higher recognition accuracy on the Newt, but it's
carried over into my day-to-day writing. It's interesting to see how
utterly unreadable something I wrote just two years ago is compared to
what I write now.
My point is that I adapted to my Newton's recognizer faster than it
adapted to me. I would bet that the Tablet PC's HWR engine would be
used in a similar way. Plus, imagine how handwriting would improve if
everyone got one! Well, either that or profanity use would increase.
The second one.
On a side note, I absolutely LOVE Microsoft's new Office program,
OneNote. It's almost EXACTLY like Notes from the Newton OS, but without
outlines. It probably has them somewhere, but I'm too lazy to find
them. It recognizes ink as drawings even without the HWR engine
installed. With the HWR, you can just write into it and have it
recognize stuff as text. It even has recording! Very nice program.
The only thing I would really change about the Microsoft HWR engine is
that I would first run the ink through a gesture recognizer. The
Newton's gestures make many tasks SO MUCH EASIER and that was
extraordinarily stupid of Microsoft not to include something like that
in Windows xp Tablet Edition. As it is, I have to use StrokeIt for my
mouse gestures and it doesn't work so well with a tablet.
-- Dylan Stewart AC5ZH -- This is the NewtonTalk list - http://www.newtontalk.net/ for all inquiries List FAQ/Etiquette/Terms: http://www.newtontalk.net/faq.html Official Newton FAQ: http://www.chuma.org/newton/faq/
This archive was generated by hypermail 2.1.5 : Mon Feb 16 2004 - 20:00:02 PST