22 Jan
2022
22 Jan
'22
7:02 a.m.
Has anyone ever used Amazon Mechanical Turk to employ typists to type in
old listings of lost code?
Asking for a friend.
5:57 p.m.
No, OCR totally fails on olde line printer listing. At least the ones I've
tried (tesseract, online, ...)
On Sat, Jan 22, 2022 at 8:06 PM Ethan O'Toole <ethan at 757.org> wrote:
Can the listings be OCR'ed?
- Ethan
Has anyone ever used Amazon Mechanical Turk to
employ typists to type in
old listings of lost code?
Asking for a friend.
--
: Ethan O'Toole
23 Jan
23 Jan
7:11 a.m.
I've run into that situation too, with listings so difficult that even a commercial
OCR program (FineReader) couldn't handle it. At the time Tesseract was far less
capable, though I haven't tried it recently to see if that has changed.
Anyway, my experience was that the task was hard enough that it needed someone with
knowledge of the material. It may be a contract typist could do a tolerable job but I
have my doubts. Typing, say, an obsolete assembly language program if you see it merely
as a random collection of characters is going to produce more errors than if the person
doing the typing actually understands what the material means.
One consideration is the effort required to repair transcription errors. Those that
produce syntax errors aren't such an issue; those that pass the assembler or compiler
but result in bugs (say, a mistyped register number) are harder to find.
paul
On Jan 22, 2022, at 8:57 PM, Mark Kahrs via cctalk
<cctalk at classiccmp.org> wrote:
No, OCR totally fails on olde line printer listing. At least the ones I've
tried (tesseract, online, ...)
On Sat, Jan 22, 2022 at 8:06 PM Ethan O'Toole <ethan at 757.org> wrote:
>
> Can the listings be OCR'ed?
>
> - Ethan
>
>
>> Has anyone ever used Amazon Mechanical Turk to employ typists to type in
>> old listings of lost code?
>>
>> Asking for a friend.
7:26 a.m.
I recently dealt with this with the DaJen SCI monitor listing out of the manual. The copy
is pretty bad, and either their printer was having issues, or slashing of "zero"
vs "O" was inconsistent somehow. OCRing it produced more of a mess than just
sitting with the original and a text editor open side-by-side.
I can't imagine it would've worked out well to have someone who wasn't
familiar with 8080 assembly language transcribe it, I had a rough enough time on my own,
and ended up having to compare the assembly output to a known-good ROM dump to get the
last of the discrepancies out.
Thanks,
Jonathan
??????? Original Message ???????
On Sunday, January 23rd, 2022 at 10:11, Paul Koning via cctalk <cctalk at
classiccmp.org> wrote:
I've run into that situation too, with listings so
difficult that even a commercial OCR program (FineReader) couldn't handle it. At the
time Tesseract was far less capable, though I haven't tried it recently to see if that
has changed.
Anyway, my experience was that the task was hard enough that it needed someone with
knowledge of the material. It may be a contract typist could do a tolerable job but I have
my doubts. Typing, say, an obsolete assembly language program if you see it merely as a
random collection of characters is going to produce more errors than if the person doing
the typing actually understands what the material means.
One consideration is the effort required to repair transcription errors. Those that
produce syntax errors aren't such an issue; those that pass the assembler or compiler
but result in bugs (say, a mistyped register number) are harder to find.
paul
> On Jan 22, 2022, at 8:57 PM, Mark Kahrs via cctalk cctalk at classiccmp.org wrote:
>
> No, OCR totally fails on olde line printer listing. At least the ones I've
>
> tried (tesseract, online, ...)
>
> On Sat, Jan 22, 2022 at 8:06 PM Ethan O'Toole ethan at 757.org wrote:
>
> > Can the listings be OCR'ed?
> >
> > - Ethan
> >
> >
> > > Has anyone ever used Amazon Mechanical Turk to employ typists to type in
> > >
> > > old listings of lost code?
> > >
> > > Asking for a friend.
9:09 a.m.
On Sun, Jan 23, 2022 at 9:11 AM Paul Koning via cctalk
<cctalk at classiccmp.org> wrote:
One consideration is the effort required to repair
transcription errors. Those that produce syntax errors aren't such an issue;
those that pass the assembler or compiler but result in bugs (say, a mistyped register
number) are harder to find.
You can always have it "turked" twice and compare the results.
This is also the sort of problem that modern Deep Machine Learning
will just crush. Identifying individual characters should be trivial,
you just have to figure out where the characters are first which could
also be done with ML or you could try to do it some other way (with a
really well registered scan maybe if it's all fixed-width characters).
I think if I had a whole lot of old faded greenbar etc. I would
consider manually converting a few pages then setup a Kaggle
competition for it and maybe invest a bit of money as a prize. Someone
may even have done this already (there have certainly been a number of
"OCR historical documents" competitions), but I didn't spend too much
time searching. I'm sure you're not the only one who has had this
problem to solve.
10:16 a.m.
On Jan 23, 2022, at 12:09 PM, Gavin Scott <gavin at
learn.bio> wrote:
On Sun, Jan 23, 2022 at 9:11 AM Paul Koning via cctalk
<cctalk at classiccmp.org> wrote:
Maybe. But OCR programs have had learning features for decades. I've spent quite a
lot of time in FineReader learning mode. Material produced on a moderate-quality
typewriter, like the CDC 6600 wire lists on Bitsavers, can be handled tolerably well.
Especially with post-processing that knows what the text patterns should be and converts
common misreadings to what they should be. But the listings I mentioned before were
entirely unmanageable even after a lot of "learning mode" effort. An annoying
wrinkle was that I wasn't dealing with greenbar but rather with Dutch line printer
paper that has every other line marked with 5 thin horizontal lines, almost like music
score paper. Faded printout with a worn ribbon on a substrate like that is a challenge
even for human eyeballs, and all the "machine learning" hype can't conceal
the fact that no machine can come anywhere close to a human for dealing with image
recognition under tough conditions.
That said, if you have access to a particularly good OCR, it can't hurt to spend a few
hours trying to make it cope with the source material in question. But be prepared for
disappointment.
paul
One consideration is the effort required to
repair transcription errors. Those that produce syntax errors aren't such an issue;
those that pass the assembler or compiler but result in bugs (say, a mistyped register
number) are harder to find.
You can always have it "turked" twice and compare the results.
This is also the sort of problem that modern Deep Machine Learning
will just crush. Identifying individual characters should be trivial,
you just have to figure out where the characters are first which could
also be done with ML or you could try to do it some other way (with a
really well registered scan maybe if it's all fixed-width characters). 
11:47 a.m.
On 1/23/22 10:16, Paul Koning via cctalk wrote:
Maybe. But OCR programs have had learning features
for decades. I've spent quite a lot of time in FineReader learning mode. Material
produced on a moderate-quality typewriter, like the CDC 6600 wire lists on Bitsavers, can
be handled tolerably well. Especially with post-processing that knows what the text
patterns should be and converts common misreadings to what they should be. But the
listings I mentioned before were entirely unmanageable even after a lot of "learning
mode" effort. An annoying wrinkle was that I wasn't dealing with greenbar but
rather with Dutch line printer paper that has every other line marked with 5 thin
horizontal lines, almost like music score paper. Faded printout with a worn ribbon on a
substrate like that is a challenge even for human eyeballs, and all the "machine
learning" hype can't conceal the fact that no machine can come anywhere close to
a human for dealing with image recognition under tough conditions.
The problem is that OCR needs to be 100% accuracy for many purposes.
Much short of that requires that the result be inspected by hand
line-by-line with the knowledge of what makes sense. Mistaking a
single fuzzy 8 for a 6 or a 3, for example can render code inoperative
with a very difficult to locate bug. Perhaps an AI might be programmed
to separate out the nonsense typos.
Old high-speed line printers weren't always wonderful with timing the
hammer strikes. I recall some nearly impossible to read Univac 1108
engineering documents, printed on a drum printer. Gave me headaches.
At least that's my take.
--Chuck
24 Jan
24 Jan
2:57 p.m.
On 2022-01-23 12:47 p.m., Chuck Guzis via cctalk wrote:
On 1/23/22 10:16, Paul Koning via cctalk wrote:
Document source is also a problem.
You would want to keep scan it at the best data format,
not something in a lossey format.
Ben.
Maybe. But OCR programs have had learning
features for decades. I've spent quite a lot of time in FineReader learning mode.
Material produced on a moderate-quality typewriter, like the CDC 6600 wire lists on
Bitsavers, can be handled tolerably well. Especially with post-processing that knows what
the text patterns should be and converts common misreadings to what they should be. But
the listings I mentioned before were entirely unmanageable even after a lot of
"learning mode" effort. An annoying wrinkle was that I wasn't dealing with
greenbar but rather with Dutch line printer paper that has every other line marked with 5
thin horizontal lines, almost like music score paper. Faded printout with a worn ribbon
on a substrate like that is a challenge even for human eyeballs, and all the "machine
learning" hype can't conceal the fact that no machine can come anywhere close to
a human for dealing with image recognition under tough conditions.
The problem is that OCR needs to be 100% accuracy for many purposes.
Much short of that requires that the result be inspected by hand
line-by-line with the knowledge of what makes sense. Mistaking a
single fuzzy 8 for a 6 or a 3, for example can render code inoperative
with a very difficult to locate bug. Perhaps an AI might be programmed
to separate out the nonsense typos.
Old high-speed line printers weren't always wonderful with timing the
hammer strikes. I recall some nearly impossible to read Univac 1108
engineering documents, printed on a drum printer. Gave me headaches.
At least that's my take.
--Chuck
4:28 p.m.
On Jan 24, 2022, at 5:57 PM, ben via cctalk <cctalk
at classiccmp.org> wrote:
That's true generally. Anything other than actual photographs (continuous tone
images) should NOT be run through JPEG because JPEG is not intended for, and unfit for,
anything else. Printouts, line drawings, and anything else with crisp edges between dark
and light will be messed up by JPEG. PNG and TIFF are examples of appropriate compression
schemes.
paul
...
Document source is also a problem.
You would want to keep scan it at the best data format,
not something in a lossey format. 
5:10 p.m.
That's true generally. Anything other than actual
photographs
(continuous tone images) should NOT be run through JPEG because JPEG
is not intended for, and unfit for, anything else. Printouts, line
drawings, and anything else with crisp edges between dark and light
will be messed up by JPEG. PNG and TIFF are examples of appropriate
compression schemes.
TIFF actually isn't a compression scheme, it's a tagged file format, and
one _can_ specify jpeg compression of images in a TIFF file.
Perhaps it would be better to say one should avoid _lossy_ compression
schemes on scans with crisp edges or large areas of solid color. These
are areas where jpeg will add visible noise.
De
5:54 p.m.
I've tried to OCR old Fortran Code from DTIC pdf documents.? There were
2 big problems;
1. The copies are very poor to start with and all OCR attempts produced
about 75% error rate.
2. Old Fortran code limited variable names to 6 characters so they were
generally not descriptive of what they represented.? Some characters in
the Fortran variables sometimes were missing in the printout and made
recovery nearly impossible.
I hope that now gov't contracts require code to be archived
electronically for posterity.? Probably never happen.
Doug
On 1/22/2022 8:06 PM, Ethan O'Toole via cctalk wrote:
Can the listings be OCR'ed?
??????????? - Ethan
Has anyone ever used Amazon Mechanical Turk to
employ typists to type in
old listings of lost code?
Asking for a friend.
--
: Ethan O'Toole
1063
days inactive
1066
days old
11 comments
9 participants
participants (9)
-
bfranchuk＠jetnet.ab.ca -
cclist＠sydex.com -
dj.taylor4＠comcast.net -
drb＠msu.edu -
ethan＠757.org -
gavin＠learn.bio -
lists＠glitchwrks.com -
mark.kahrs＠gmail.com -
paulkoning＠comcast.net