For the final project I am collaborating with Yonatan. So far, Yoni has been focusing on the language-processing and text-generation aspect of the project, and I have been focusing on the physical installation and Optical Character Recognition elements.
I have been doing a lot of research this past week on various OCR software. Unfortunately, so far as I have been able to find, the only software that offers not only OCR, but word and character coordinates as well (which is essential for this project) are OmniPage and TypeReader. However, both of these are proprietary software. OmniPage is $500, and I have a feeling that TypeReader is very expensive too.
Tesseract is a free, open-source OCR software. I installed and ran it, and the results are pretty good. Here is the article we have been using as a test, its from the New York Times a few weeks ago, ironically an article about OCR:
And here is the output from the sample article image file:
A Speech Lost in Digital Translation
It was Nov. 2, 1859. The jury in
Charles Town, Va., deliberated only 45
minutes before sentencing John Brown
to death by hanging for leading a raid
on the federal arsenal at nearby Har-
per‚Äôs Ferry. The court asked Brown, a
gaunt, craggy-faced abolitionist, if he
wished to speak. He did.
His defiant yet humble words trans-
formed Brown, notorious for the grisly
‚ÄòTill the c *lay they
C:IIII to me,‚Äù John
Brown said. Or did he?
murders of pro-slavery settlers in Kan-
sas, into a martyr for the anti-slavery
He sought no mercy and accepted his
fate: ‚ÄúIf it is deemed necessary that I
should forfeit my life,‚Äù he said, and
‚Äúmingle my blood‚Äù with ‚Äúthe blood of
millions in this slave country whose
rights are disregarded by wicked, cruel
and unjust enactments, I say, let it be
The speech was a bit over 500 words
long. This newspaper printed it in its
entirety on Nov. 3, 1859.
For more than 140 years afterward,
the original story was available only to
those with access to the hard copy
newspaper or its microfilm equivalent.
VmG¬•N1 it itb’aELL1oN.
-701111 Bffrwu sentenced to ncnm.
Till Addressjo the Court and Jury.
EDWARD GOPPIE F0lJND GIIILTIG
Cunuuowu, Wednesday, Nov. 2.
Messrs. Rneszu. and Sxmum-, from Boston,
reaolxed licrc to-day.
Cont was brouglit boforo the Mngisim4u’s Courh
and waived an examination.
Cor-rxz’¬ª HIM was resumed. No wilnesms Wow
called for the defence.
Hr. llnnino opened for NW C¬∞‚Äú‚Äú‚Äú¬∞‚Äú”¬∞‚Äú3″‘ 9
Messrs. Ho:-rand Gnnwow followall for R19 ¬¢14f0Ni~
un, and Mr. ilunuu nlosed for the prowwtlfm- ‘I`h¬¢
speeches were of marked ability.
‘ Kr. GIYIFGLD asked for several Instructions to the
jury, which were all granted by the Court: W5 ‚ÄòM
jqry retired. `
Bums was then brought In, and the Court/’im
was Immediately ituongod. ¬ª
The Umm gave its decision on thc motion lor rm ‚Äò
arrest of judgment, overrnling the obgenticms made.
Qu the objection that treason cannot be ocmmttied
f ` ‘ng n¬ª1a.14b¬ª*w%%% ai\cc
HEAR YE On Nov. 3, 1859, The New
York Times printed a defiant speech
by the abolitionist John Brown.
Then, in 2004, The New York Times be-
gan the process of making its archive
into a searchable database. In 2008, it
hired Pittsburgh-based reCaptcha to fix
the errors. This project is expected to be
completed by the end of this year.
Brown‚Äôs closing speech can be found
as it originally appeared in The Times
by doing an advanced search at
nytimes.com, using the headline ‚ÄúVir-
ginia Rebellion. John Brown Sentenced
Close examination of this image
shows subtle unevenness in the inking,
differences in how a particular letter
looks from word to word, some
smudges and occasional wobbles in in-
dividual words or letters.
Despite these shortcomings, the hu-
man eye can easily read the text. The
optical character recognition software,
or O.C.R., however, had a terrible time.
Robert Larson, The Times‚Äôs vice presi-
dent for search products, said parts of it
looked like ‚Äúsomebody swearing in a
The last few lines of Brown‚Äôs speech,
for example, were reproduced by
O.C.R. as follows: ‚ÄúI fear it lhas becn
stated by Of 1 bave to join nIO, but tlhe
is true I do not say this to them, but as
regretting their -.. Notone but joined me
of his on, ltc* ord, and the greater part
at expense. Number of them I never
saw , and never hld a ot conversation
with till the c *lay they C:IIII to me, and
that for the I halve stated. Novw I have
For human eyes, however, the pas-
sage is child‚Äôs play. Computers may be
able to win at ‚ÄúJeopardy!”, but on this
stuff they‚Äôre about as clever as marmo-
sets. GUY GUGLIOTTA