The Online Genealogist

John Brugliera

OCR = Occasionally Correct Reader

When you search within scanned newspapers or books online, the results are usually derived by optical character recognition software, more familiarly known as OCR.  The idea is for the software to “read” the printed characters and “translate” them into readable/searchable words and sentences.  Unfortunately, that simple-sounding task isn’t so given the inconsistent quality of the printed and scanned material – especially when it comes to older newspapers.

Here’s more info than you’ll ever need on OCR…

http://www.dlib.org/dlib/march09/holley/03holley.html

And a great example of how newspaper print can “confuse” the software…

Image

Yes, very iffy and not fully reliable.  “At” comes up as “la”, “the” is “che” and “good” here is read as “gobd”. But then it actually reads “lovers”, which looks more like “lpvera” in the copy.  As you can see, OCR resluts can be extremely sporadic.

When words are hyphenated, OCR gets really confused, which can be a source of “amuse*ment” for some, but frustration for most.

How well do you think OCR can read this obituary page from The Sun (New York City)?

Image

Maybe it will properly capture that bold “Mary” about halfway down, but the rest of it?  Well, let’s see, shall we?

There are several free OCR readers online; just upload your image and see it magically transformed into perfectly readable text!

http://www.free-ocr.com/

Uhhhhhhhhhhhhhh, maybe not…

||;1. – I’u.z l.. ;\..-»-. f.;,,| , ;»,
“”~ ‘1 “8 K ‘–‘ 1. 3. “‘1 -ll IP12
’14‘ :|..n !~»_’: 4. ‘-. – ‘ .‘ . ‘H “°u~
“I 1;? N_lv»\TNS0¢ Y‘ … in ~-‘ bl.‘ 4 u-I Tm ,’-I :1 , J01; I’, pl 1′ P,
-» “‘|W.‘- 9’ P‘ ‘1 H” – ~ ‘~ u “I, or \‘4I|-0“ 0 , 1I“\
“””‘*”‘ ” M – . W ‘  1. 4) -‘,‘-x~-A-.. 0’\O’ luau
V‘ ‘6-‘ “. ‘I “W I ” 5 \o.:”‘ ‘film llgl. ‘ ll ‘1 l|\ .1 |f0ug;¢
I’M.‘ -0.”. L‘ -0″‘ , 44 Q uflh-,-fp 4|. ‘nu ‘*0 1| \..;§’ ‘ .’f.».
: rt E ‘u“ “UV 01.0”] ’00 1.1‘   ‘In fl,§|| “||‘h’n l|.0’Jf‘¢
|_|’ 1′ K —\”\ l’r ol’\\’ 1.!” , “l.| ‘ IN»-‘I| H;-H K ¢|v|||,g||.
“r;.\-V | , » .!l . ‘ ‘\, I‘, I o_ .‘ .  .\.o,‘f-,.0 |~‘u“lf,“‘_“|‘
‘ 0|-I110,‘ ||”\’;-0. ‘ad Yf,v9¢I’ .|’ O7″. “9u”.‘v. “.n |…\ ‘um.
u ‘K’.’ ‘ ‘1? -‘ x f‘ .  ” *‘?”‘ ‘\ “ -‘“ ‘ *–‘””*|
. ..“‘||’H\‘ .\\~ |¢0,,‘¢‘“’ ” L.__ ‘Q f qg.‘ u|”.o,,| ‘“|’~ P”… ..
‘ Tm. ‘l.’::r |-1 ‘§l,.‘w” WW” ““‘ ‘ “””°”-“lam N~’ml¢!-
“‘|”l-;~:;§\‘- _’_i1;! I.» H. rm |>-1-\.g.y,’.MI|. O, ION, Am“;
Th H1: rn \;= :1 V? “”1 H vw ‘ ~w.v’
‘ \ .. , ‘ ‘ . ‘ ‘ .0 0 \ O
M ‘xi H) W W,“ Jar’ _4‘. _: 1:. -(‘$10! rntchluc NI
x  \’\‘.‘l.,’:’c t | ‘H1 \ .‘.\. _.l_“‘_‘  .”0“”|.Q “0”
|:\tn|’|:..1vh~’\.uh” “T. .3 ‘hnurw M‘. “”, L‘
‘ – – ~ “ll” — 1 ad uln-
‘ f~’.~*f’||.\\.:}_“~-4 !.!”»’I»~w: ‘.~’ _H’: l»\uvu_\ .m- rm-|m~Iln;|y
-b ’ I-I n’t‘H\‘| 0″. “U ‘9‘! ‘8. ‘“w\ ‘mM..m..”‘ M.’
:»|r~”;’|:~-Q-|’;v.’.|n|: ‘1 LVN” .l:’\l. I ‘H’. N“ “m“m““”
s é 1 ‘, ‘ ‘
:‘|.\H’;:|\ – “ll Ihu”°»H\ -‘1” 1.1%‘, Huh!“ “.§|H\Ih
Q :~’:r:rI -u|)\1u\t‘I~bl1. \ –mm luau. Inland. llvd ll
| “u”_””fff!;” I_’=’l {‘4 I1-if fir» rmvct!ull{ lam.-Mo
bu?‘ “ f 1-‘.|.a~._§_‘._. Rf!“.I§:-“rr‘1al’°fl\°o-‘:0! uh r-mmlg,
; “NJ” M.‘ .’ ” ‘; Q. IMP |Hu.~ “0.-I LT: r.!I;.t’|’l.:a.o.:l.|.f¢?~.|‘Us’:
y ‘\’\”:”3 a’~u7″‘|’ “”£1r1′;”‘””‘»’::“;~?’;”:|”‘“:”;.”§’
‘ A 0. Q ‘I |-‘Q . * \ .  5“ .  . ~. .
‘ ‘MS: ; ‘rE._’¢“. -‘| HI‘o’r|’.-Hf.‘-“1′.\.m bu c”
i \’“|’|»” \\|\ In; \..;ln’ IA! I-n I nfln l\-nL_-l..-

Wow, it didn’t even pick up “Mary” as I thought.  The only word I see is “Inland”, which is supposed to be “Ireland”.  (Funny that WordPress itself translates some of the text into smileys!)

Let’s try another image.  It’s darker and crisper, so the results should be better than the above.

Image

I highlighted two couples — the thinking is that the top pair will be read better than the bottom.

IA It III”. –
IIIQWK-onllllzl-l..-In , Wm; 9
no-I».l at st. A|}n’n on the |\’.-an-. LV”.I\e- ll»-v..’8?nl‘
mu firhnurl. I.II.. rmvfin OHQM Iv-own us-In it
£_li:’:0:’ou“ .¢::’,:nvt 0! JOUII I. lid NM D10 null .
H ‘ , , I ‘ .
,,.’*.3″.::*::’>:;.i”:.’;’¢..:’:.‘!”;’-.» n~.r.”.”-:a’t:-
IL I ‘ 0 ‘ ‘ ~ O ‘ I Q
no-1|-~r|» .\. I Mm! M Um 1-an to Marv Jo‘. cl mgmur of
{Iv I:-tr J. lhnord 510001001! Nllllvlllj‘. Him.
. ‘ ‘~‘.“’.‘3’»“ Y I i|‘»’2i=§’2‘-“i’8#’; ‘6’?» °’.~2!2..%‘.’.‘€.‘
 0″ |, ‘ N‘ :\’ – ‘ . P 0 0 ‘ O
‘ 1′ I.a‘tt ‘oi  1;. 5′ ~’.‘\u‘3illI B-. dllfllw-I’ 0| U00 UM
‘ ,1 ~‘ ° | ‘ U 0 I. .
| “1§|||’l‘|\’.-s – iwrnnvnn -In Joohonvllln. Hm. ol
.\Iua0n‘\’. Jan. M ny W. N. l|\\r3m%Ju0ucc- e.l we Puke.
–.’.» :;~.’ £ :’.’2″‘<:.’=”‘~.: m-2:. ~ “”‘.’;:-“‘.:~.,……
‘ § . ‘“ 0 “
Jan. :1, II |;mlwr|w|Pl’n|-K. ilk; I§tbe IlrV._ L‘:
llamlltmml. “’o|l¢’I’ Imug|;m of ‘WI ufk In (.lI’fI”‘
daugmer of ca» nu» Iv. s nude Inn or llw II-c
-I .
ul’ll.l. J \f‘IlSn\i (m Tm-ndny. P0701, by Oh! RN‘
mu. rm»… |~. 1» , \\m um to Mu-e, ammo: oi
w(“|’vi’l1′?‘*:i’v”‘ M!|””‘ ‘i’r’|”|\r,<‘»?» “.1950 Wvdnflbl
, , \ . . ,
r\’mln~’. I» » n. I. an nw nrmlnuci umhv lm¢IQ’0 (vii;
Mr \ Ilmulu” Hhmmlnl annmgn Flvrllqn. ‘ ..l0
Mn» .\~lq|~: l.. lfznm, wlnnlfierof Ir. Wm. la. ll nu.

But then again, maybe not.  No wonder this OCR is FREE.  (Yes, some OCR is better than others.)

If it can’t read THIS image, then this OCR is basically USELESS…

Image

With regard to online newspaper image quality, this is about as good as it gets.  The OCR results…

IA ll BIRD.
luflsnl-BEBKIRBAIYOII.–On T0009“. his 15
at Cmlla Church. b tho llwv. Ila I Iv . hal-
Iof. htolnll ;. Ill to OIIMI cl! 00.3″: gt!!!»
W I –
Damn Print“! 1‘ umuiwlnl .l..
vi .na~ A RIB.-On “mum. Jun 10. at we
“nee of gm bride‘ rum. by uvkv. Mr. ‘nu,
r. lllmn oven 03 gflllllllkg. 1% Mm mu
“DER, magma: 0 . . an. .. ol claim
11-v-
DI V0 ICED.
LA IZY.—In the my ol Rex‘ York. Ml QM 101,?
Q I | 0 b’ 0 J
::::§*‘:.’%:.z.::'”,’.:a: .2’l..’.:., ..’:r:’u…”:’m.*.,’.
 -ani
DIED.
lIIU.—0| J I
€:%:c|’r:ac,nuct or gear) in
  21:2:
n mmnv. one 81.01 crazck I’. ll. lhlouvon 0|‘
Q an rrgccllulznot to uugmt.
…~.:<:>;.w.~-::…:.; “-2.1% .~.::~”- “–‘–
0 nl0tlvr‘_na¢: 1110010.” the $inu|’y. and than 0|
In non. Jun lhmou. on rnloootlnuy mange! co nt-
an In Marni Iron: Mr mo mudnoce. 810 cot IMO
at on goo-lay. Juno Ital, I’. I.
I10! \l0. 0 anmlay. Mm tn. Cnherlnt In-Ion!
1).‘: fncboluqnrnm. to the lmh {far M er cm.
R ouvn and rlentotuc um lumd I
0¢uod*no hum] Irom r Iota Fianna‘. III Mount
IL, II u1’ncQ‘. inn II. it 90′: net A .,0la0nn- 0|
1‘; .{“.’!’.’_‘! L_’!¢P;_¢”t’!’-“.9!.L’¢_9.t1.wLlMm~ rm

Wow, is that bad or WHAT?!?

Going back to the original example of the bad OCR, can it even read the RIGHT SIDE correctly??

‘nu ‘h:u?””?-.“’ ‘ ‘ uh I Family Theatre.ThU theatre wli open to-day withan
1 U48 ‘ . . . .
u “Mn ch…“ u H” n‘ I. ‘uh ent|re change of b|Il at Its mat|nee.The bull la of gobd
,¢g_ ‘pp 5‘ |i_ Q 3*] ygflgg “Q variety andwill surely please all lovers of amuse”ment.
V.” WNW 9’91 I” |F””@ 3.3.‘ One of che feature is’ MissMinerva Vano. \ h e queen
nun; ‘Om- at am fawn U’ lb .
‘mar’ ‘an.’ “t Q…‘ 0‘ “. ~‘.‘. of the handcuffs.We copy the follow|ng from
‘ml; we can the loltou-In hon
|’|’||Q flung.‘ |4»“¢;_ Rig Ilgqgg, The Evening Leader. New Haven,
|¢oan: .
Conn:

Better, but still unusable.

The bottom line here is that OCR is NOT at the point where it’s fully or even partially reliable in most cases, ESPECIALLY when it comes to newspapers.  Should you ditch the OCR searches altogether?  No.  But just realize that if you don’t get the expected results, it does NOT mean that what you’re looking for is NOT within that book or newspaper.  You’ll just need to browse the pages on your own, using your OWN built-in OCR.  And as shown above, even THAT can be difficult given the varying quality of original printed matter.

 

And if you’re looking for a human OCR, I’m your man!

 

Image

 

Advertisements

Single Post Navigation

2 thoughts on “OCR = Occasionally Correct Reader

  1. Pingback: It’s a Mocavo Two-fer! | The Online Genealogist

  2. Thanks for finally writinmg about >OCR = Occasionally Correct Reader | The Online Genealogist <Liked it!

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s

%d bloggers like this: