On 27 October 2011 22:24, <Bowerbird@aol.com> wrote:

jimmy said:

...

   they did quite a lot of work to bring the engine itself up    to modern standards, and have been relatively successful

jimmy, that's fantastic news!

i myself don't do too much scanning, not books -- more interested in _birthing_ digital-books -- but the people over at d.p. will love to hear this...

they've really tried to use tesseract in the past -- and i mean they truly _wanted_ it to work! -- but it just didn't perform at the level required...

Yeah, I saw one of the test projects. That was version 2.

so they keep using finereader.

(and, for anyone out there wondering about it, no, they don't use the o.c.r. from archive.org... even when archive.org stuff isn't fatally flawed, the d.p. people have found they can usually get better results from their own scanning efforts. don't know why that's true, but it's interesting.)

IIRC, there are other reasons, such as the way the prep tools work.

so i hope d.p. people from here spread the word.

(they usually erect a big cone of silence around this listserve, because they don't want the people from over there to know about me, or hear me... they banned me over there, where i'm now known as he-who-cannot-be-named, a.k.a., voldemart.)

...

   It got a huge shot in the arm    when the Android guys took notice

ok, that makes sense.

because i've definitely noticed that it's now being incorporated even at the level of apps these days.

and i must be honest that one thing which i had never anticipated is how it would be paired with _translation_ software, for some _killer_ synergy.

my word!

those apps, where you point your phone at a sign and it is translated in real-time, are stupendous!

(MT happens to be my main area, so it's a development I'm particularly interested in :) The first Android work with Tesseract was on tools for the blind, for reading street signs aloud. The text-to-speech component from the same group was recently added to Android and Chrome at a general level, so that group are probably a good indicator of what will be happening in 2-3 years time.

and when you think how they could be improved with some crowd-sourcing and mechanical-turk, the realm of possibilities gets truly staggering!

Yes, I know a few people who are working on this general area from different angles. (Plus, quite a lot of the material that the big players used came from software localisation for open source projects, so in a sense, they've been using crowd sourcing for several years now).

...

   is due to be released around Thanksgiving?    I'm not American, so I only have a vague idea    of when that is :)

the 4th thursday in november.  4 weeks from today.

Ah. I'll put it in my calendar :) -- <Sefam> Are any of the mentors around? <jimregan> yes, they're the ones trolling you