why hasn't "old" shorthand made a comeback with the advent of OCR?

[Gabriel Holmes]

...and could it?

[Matt Thomas]

The palm pilots had a stylus shorthand since the 90's that depended on the stroke order for recognition.

Sent in Transit (mst)

---

From: Gabriel Holmes <holm...@gmail.com>

Sender: plove...@googlegroups.com

Date: Mon, 12 Nov 2012 11:40:46 -0800 (PST)

To: <plove...@googlegroups.com>

ReplyTo: plove...@googlegroups.com

Subject: why hasn't "old" shorthand made a comeback with the advent of OCR?

...and could it?

[Tony Wright]

I think the two biggest obstacles are:

(1) the number of conflicts (i.e., one stenographic symbol
representing two or more words or phrases) in handwritten steno, just
like with pre-computer stenograph theories. It was easy for humans to
disambiguate these, but many would be a tough problem for computers.

(2) I know that, at least with Gregg, many letters are disambiguated
by length, so that the difference between T and D is that T is a short
upward stroke, and D is a long upward stroke. Same principle for S vs.
F vs. V, and many others. I think a computer program would have to
learn the relative lengths for an individual writer, which would vary
widely from individual to individual.

So I think a computer-readable handwritten steno would require
something other than Gregg, at least.

--Tony

On Mon, Nov 12, 2012 at 1:40 PM, Gabriel Holmes <holm...@gmail.com> wrote:
> ...and could it?

[Gabriel Holmes]

I was thinking, maybe if there were something like Swype for plover? Instead of a chord, you'd have a curve passing through all the keys? That ought to be something fairly doable, with the touch screens...

On Monday, November 12, 2012 2:40:47 PM UTC-5, Gabriel Holmes wrote:

...and could it?

[Smirnow, Krzysztof]

Well, I was thinking about it a lot. Now I think, that it could be possible - if these mathematicians, which work on for example voice-to-text recognition, use their inventions to make shorthand-to-text recognition. The difficulty-level is similar, however shorthand should be easier. But nobody is interested now in this way of input data to computers, so, perhaps, it will die in oblivion.

2012/11/13 Gabriel Holmes <holm...@gmail.com>

I was thinking, maybe if there were something like Swype for plover? Instead of a chord, you'd have a curve passing through all the keys? That ought to be something fairly doable, with the touch screens...

On Monday, November 12, 2012 2:40:47 PM UTC-5, Gabriel Holmes wrote:

...and could it?

--
Krzysztof Smirnow.
flame...@stenografia.pl
http://www.stenografia.pl
--| Nondum lingua suum, dextra peregit opus |--

[Zerk]

I thought of something like that for thumbs on a phone screen, but the whole reason chords are faster is that you're using several fingers at once. I'm pretty sure that Swype could be optimized by using a different layout than qwerty, but I don't think it would be an order-independent one.

[Gabriel Holmes]

I recently came across something called 8pen, and I'd love to see a version of this for Gregg shorthand...

[Jack]

On Monday, November 12, 2012 1:40:47 PM UTC-6, Gabriel Holmes wrote:

...and could it?

From what I know of Gregg, no, it couldn't, at least not at it is now. Even today, OCR requires manual auditing and correction by data entry workers even given uniform digital font faces. Imagine trying to OCR cursive writing. This is basically what you would be doing if you fed Gregg to OCR software. Everything is relative. Big circle, little circle; long line, short line, medium line—this seems hard to pull off unless the OCR software is intelligent and can define per-session understanding of each stroke relative to the others.

 

[Jeremy Kahn]

The intelligence you describe is already at work in smart phone "Swype" keyboards and of course I'm handwriting recognition systems. We - as a civilization - have the technology, but I suspect no one's built it for shorthand because there's no money in it.

Capitalism, man. Ruins all my fun.

Jeremy

[Jeremy Kahn]

And in some kind of Murphy's Law Enforcement, Swype let me down and wrote "I'm" where I meant "in". 

Point is, context awareness is what makes speech and handwriting recognition systems work today, and they work particularly well for the same places you'd want to use shorthand: with a single "speaker" (transcriber), recording near-continuous speech or text (so that adjacent words provide extra clues to decoding).

There aren't AI blockers here. There are time and money blockers, alas. But the ASR and handwriting technologists know what would work.

Perhaps a graduate student would be interested in training a shorthand recognizer, if a suitable collection of shorthand (plus decodings into English?) were available. Know of any such collection? I know some keen electrical engineers!

Jeremy

[Gabriel Holmes]

the symbols in Gregg shorthand are all taken from a single pattern -- an oval with a cross in the middle. Maybe the input section could just be something like that, and you could lift your finger between strokes.Then there wouldn't even be a need for any sort of OCR.