/LMAR=0/XLINE=3/FONT#0=BAXL30/FONT#1=BAXM30/FONT#2=BASB30/FONT#3=SUB/FONT#4=SUP/FONT#5=BASL35/FONT#6=NGR25/FONT#7=MATH30/FONT#8=FIX25/FONT#9=GRKB30
␈↓ α∧␈↓␈↓ u1


␈↓ α∧␈↓α␈↓ α⎇OPTICAL CHARACTER READING IS STILL AN UNSOLVED PROBLEM

␈↓ α∧␈↓␈↓ αTFor␈α⊂many␈α⊂important␈α⊃applications␈α⊂optical␈α⊂character␈α⊂reading␈α⊃is␈α⊂still␈α⊂an␈α⊃unsolved␈α⊂problem.
␈↓ α∧␈↓These␈α
include␈α
the␈αreading␈α
of␈α
books,␈αmagazines␈α
and␈α
newspapers␈αin␈α
uncontrolled␈α
fonts.␈α
 With␈αthe
␈↓ α∧␈↓advent␈αof␈αthe␈αnewest␈αgeneration␈αof␈αdisk␈αfiles,␈αtypified␈αby␈αthe␈αIBM␈α3830␈αthat␈αstores␈α2.52␈αgigabytes
␈↓ α∧␈↓per␈α∞unit,␈α
it␈α∞has␈α
become␈α∞economical␈α∞to␈α
make␈α∞a␈α
national␈α∞computer-based␈α
library␈α∞that␈α∞would␈α
store
␈↓ α∧␈↓everything␈α
that␈αhas␈α
ever␈αbeen␈α
printed␈αand␈α
make␈αits␈α
text␈αavailable␈α
from␈αany␈α
suitable␈α
terminal␈αin
␈↓ α∧␈↓the␈α∞country␈α
over␈α∞the␈α
telephone␈α∞lines.␈α
 Besides␈α∞this,␈α
social␈α∞science␈α
research␈α∞requires␈α
the␈α∞ability␈α
to
␈↓ α∧␈↓read␈α⊃large␈α⊃quantities␈α⊃of␈α⊃text␈α⊃into␈α⊃computers␈α⊃for␈α⊃content␈α⊃analysis,␈α⊃and␈α⊃surely␈α⊃the␈α∩CIA,␈α⊃whose
␈↓ α∧␈↓function requires monitoring thousands of foreign publications could use the capacity.

␈↓ α∧␈↓␈↓ αTThe history of the problem is as follows:

␈↓ α∧␈↓␈↓ αTIn␈α
the␈α
1950s␈α
there␈α
was␈α
much␈α∞research␈α
in␈α
optical␈α
character␈α
recognition␈α
supported␈α∞by␈α
many
␈↓ α∧␈↓government␈α
agencies.␈α
 Some␈α
promising␈α
results␈α∞were␈α
obtained,␈α
and␈α
commercial␈α
systems␈α∞that␈α
could
␈↓ α∧␈↓economically␈α
read␈α
special␈α
"ocr␈α
fonts"␈α
were␈α
developed.␈α
 These␈α
machines␈α
have␈α
been␈α
in␈α
widespread
␈↓ α∧␈↓use␈α∞since␈α∞that␈α∞time,␈α∂but␈α∞depend␈α∞entirely␈α∞on␈α∂the␈α∞fact␈α∞that␈α∞the␈α∂fonts␈α∞and␈α∞the␈α∞manner␈α∂of␈α∞printing
␈↓ α∧␈↓them are designed with OCR in mind.

␈↓ α∧␈↓␈↓ αTSome␈α∞time␈α∞in␈α∞the␈α∞early␈α∞1960s,␈α∞government␈α∞research␈α∞support␈α∞agencies␈α∞decided␈α∞to␈α∂declare␈α∞a
␈↓ α∧␈↓victory␈α∂in␈α∂OCR␈α∂and␈α∂leave␈α∂research␈α∂to␈α∂the␈α∂companies.␈α∂ The␈α∂victory␈α∂celebration␈α∂was␈α∂premature.
␈↓ α∧␈↓␈↓↓Almost␈α
twenty␈αyears␈α
later,␈αthere␈α
is␈α
still␈αno␈α
company␈αthat␈α
will␈α
take␈αfixed␈α
price␈αcontracts␈α
to␈αread␈α
books
␈↓ α∧␈↓↓into␈αa␈αcomputer␈↓.␈α There␈αappears␈αto␈α
be␈αno␈αcurrent␈αgovernment␈αsupported␈αresearch,␈α
although␈αthere
␈↓ α∧␈↓as␈α∩recently␈α⊃been␈α∩research␈α∩in␈α⊃holographic␈α∩methods,␈α⊃supported␈α∩in␈α∩a␈α⊃search␈α∩for␈α∩applications␈α⊃of
␈↓ α∧␈↓holography.

␈↓ α∧␈↓␈↓ αTThe␈α∀main␈α∀private␈α∪research␈α∀and␈α∀development␈α∪has␈α∀been␈α∀done␈α∪since␈α∀the␈α∀late␈α∀1960s␈α∪by
␈↓ α∧␈↓Information␈α
International␈α∞Inc.␈α
of␈α
Culver␈α∞City,␈α
California.␈α
 This␈α∞resulted␈α
in␈α
the␈α∞early␈α
1970s␈α∞in␈α
a
␈↓ α∧␈↓system␈α⊂called␈α⊃Grafix␈α⊂I.␈α⊃ It␈α⊂is␈α⊃based␈α⊂on␈α⊃one␈α⊂of␈α⊃III's␈α⊂computerized␈α⊃microfilm␈α⊂readers,␈α⊃a␈α⊂Digital
␈↓ α∧␈↓Equipment␈α⊃Corporation␈α∩PDP-10␈α⊃computer␈α∩(KA-10␈α⊃vintage),␈α⊃and␈α∩a␈α⊃specially␈α∩designed␈α⊃binary
␈↓ α∧␈↓image␈α∩processor.␈α∩ The␈α∩system␈α∩includes␈α∩terminals␈α∩on␈α∩which␈α∩the␈α∩system␈α∩can␈α∩display␈α∩gray-level
␈↓ α∧␈↓images␈α
of␈αrejected␈α
characters,␈α
so␈αthat␈α
"reject␈αoperators"␈α
can␈α
key␈αthem␈α
in.␈α Two␈α
systems␈α
were␈αsold
␈↓ α∧␈↓some␈α∃years␈α∃ago,␈α∃one␈α∃to␈α∃the␈α∃U.S.␈α∃Navy␈α∃Republication␈α∃Facility,␈α∃which␈α∃uses␈α∃it␈α⊗to␈α∃republish
␈↓ α∧␈↓maintenance␈α
manuals␈αfor␈α
older␈αaircraft␈α
and␈αone␈α
to␈αthe␈α
British␈αDepartment␈α
of␈αHealth␈α
and␈αSocial
␈↓ α∧␈↓Security,␈α∂which␈α∂uses␈α∂it␈α⊂to␈α∂read␈α∂hand-printed␈α∂social␈α∂security␈α⊂forms␈α∂and␈α∂is␈α∂currently␈α⊂having␈α∂the
␈↓ α∧␈↓system␈α∂improved␈α∂so␈α∞as␈α∂to␈α∂read␈α∞the␈α∂card␈α∂catalog␈α∞of␈α∂the␈α∂British␈α∞Museum␈α∂(the␈α∂equivalent␈α∂of␈α∞our
␈↓ α∧␈↓Libary of Congress).

␈↓ α∧␈↓␈↓ αTWhile␈αGrafix␈αI␈αhas␈αread␈αmany␈αbooks␈αand␈αmagazines␈αin␈αseveral␈αlanguages␈αwith␈αsatisfactory
␈↓ α∧␈↓error␈αand␈αreject␈αrates,␈αit␈αis␈αvery␈αexpensive.␈α A␈αminimal␈αsystem␈αsells␈αfor␈α$2.5␈αmillion,␈αand␈αIII␈αis␈αnot
␈↓ α∧␈↓currently motivated to attempt a lower cost version without seeing more of a market.

␈↓ α∧␈↓␈↓ αTRecently␈α∩the␈α∩Kurzweil␈α∩Company␈α∪of␈α∩Cambridge,␈α∩Massachusetts␈α∩has␈α∩developed␈α∪a␈α∩widely
␈↓ α∧␈↓publicized␈α∞system␈α∞selling␈α∞for␈α∞$25,000␈α∞to␈α∞enable␈α
the␈α∞blind␈α∞to␈α∞read␈α∞books␈α∞in␈α∞arbitrary␈α∞fonts.␈α
 The
␈↓ α∧␈↓system␈αdepends␈αheavily␈αon␈α
the␈αability␈αof␈αthe␈αhuman␈α
to␈αguess␈αomitted␈αor␈αmistaken␈α
characters␈αand
␈↓ α∧␈↓to␈αrecognize␈αwhen␈αthe␈αmachine's␈αoutput␈αdoesn't␈αmake␈αsense␈αand␈αtry␈αagain.␈α At␈αleast␈αin␈αits␈αpresent
␈↓ α∧␈↓state, the Kurzweil machine is inadequate to read books into a computer.

␈↓ α∧␈↓␈↓ αTWe␈α∀believe␈α∀that␈α∃the␈α∀technical␈α∀problems␈α∀of␈α∃making␈α∀a␈α∀cost-effective␈α∃character␈α∀reading
␈↓ α∧␈↓␈↓ u2


␈↓ α∧␈↓attachment␈αfor␈αa␈αcomputer␈αare␈αsolvable␈αin␈αless␈αthan␈αfive␈αyears␈αwith␈αextremely␈αhigh␈αprobability␈αof
␈↓ α∧␈↓complete␈α∂success.␈α∂ However,␈α∂it␈α⊂will␈α∂require␈α∂overcoming␈α∂the␈α⊂myth␈α∂that␈α∂the␈α∂problem␈α⊂has␈α∂already
␈↓ α∧␈↓been solved and is unworthy of scientific attention.

␈↓ α∧␈↓␈↓ αTA research and development program might include the following:

␈↓ α∧␈↓␈↓ αT1.␈α
Contracts␈α
to␈α
use␈α
present␈α
equipment,␈α
with␈α
III's␈α
in␈α
house␈α
Grafix␈α
I␈α
as␈α
one␈α
obvious␈α
candidate,
␈↓ α∧␈↓to␈α
read␈α
a␈α∞substantial␈α
body␈α
of␈α∞material.␈α
 For␈α
example,␈α
keeping␈α∞Grafix␈α
I␈α
busy␈α∞for␈α
one␈α
shift␈α∞for␈α
a
␈↓ α∧␈↓year␈α⊃or␈α⊃two␈α⊃with␈α⊃technical␈α⊃material␈α⊃in␈α⊂(say)␈α⊃computer␈α⊃science␈α⊃would␈α⊃shake␈α⊃down␈α⊃the␈α⊂present
␈↓ α∧␈↓algorithms.

␈↓ α∧␈↓␈↓ αT2.␈αAnnouncement␈αof␈αa␈αprogram␈αthat␈αwould␈αbe␈αresponsive␈αto␈αunsolicited␈αproposals␈αfrom␈αthe
␈↓ α∧␈↓universities and industry.