keitai-l@appelsiini.net writes:
>Now that mainstream keitai (and peripherals thereof) can take digital
>photos, there's a potential new handwriting input modality: write
>something
>on paper, take digital photos of the page, and pass the images to an OCR
>system.
>
>Don't everybody start laughing at once.
<irrelevant>I read somewhere that the fax machine was invented in Japan
largely because of the difficulty of input and transmission of kanji via
Telex.</irrelevant>
(what follows are random thoughts rather than anything coherent)
Most of the "stuff" I need to file temporarily is wholly resistant to OCR.
(Things like the to-me-incomprehensible ingredients panel of various
shampoos in the shop so my wife can tell me which one she wants.) I
really want to create a loosely defined "heap" of things (images, audio
snippets) that I can access via voice tags. (Or keyboard tags. Voice
control of keitais has the same problem that pr0n has on keitais - the
imode handset is often useful because it creates an area in which one can
browse and RESPOND privately in a public place.)
Based on what you have suggested, but somewhat easier would be voice/sound
recognition based filing of digital images. In other words, don't bother
with OCR, but file the keitai camera images based on user defined
categories. In this case there would have to be very limited "word
recognition" ( a few verbs) and the rest is "sound recognition" (for the
categories, though some would be predefined.). So for the computer, the
problem is not "what is this word from a possible range of 70,000" but
"what of the existing categories in my memory is the idiot trying to
access". Which is a rather more accessible problem as the closest match is
likely to be right.
eg
take pic of restaurant menu. Say File: restaurants: New Pile : The Frogs
Legs: Menu" Image is filed and tagged as "menu". Then take picture of
pretty waiter/waitress. Say "File: Restaurants. The Frogs Legs: staff.
Press button for audio recording: "My name is Nortiko", Say: Save with
pile "the frogs legs". Only "Save","File" and "New Pile" are verbs.
etc etc
The problem is (obviously) memory, but that seems to be getting less and
less of an issue as time goes by. And lots of these things are for short
term filing. Anything longer term - well - if I want to keep it that much,
why not connect it to my home computer and let that do the OCR?
Handwriting is a very inefficient way of inputting a broad range of data.
Now I think about it, I don't handwrite any more even on paper. As for
Japanese kids handwriting kanji?! Ha! - They have forgotten most of it by
the time they are 23...
One place where I am surprised there has been no OCR scanning in keitai is
of meishi. An ultra slim meishi scanner should be just about feasible in a
largish keitai. In that case, the OCR would at least have a fairly clean
image to work on.
It would be nice to be able to put a tiny machine readable sticker on a
meishi as well, to make it machine readable.
Nick
Received on Sat Jan 5 14:31:07 2002