I have a lot of listings from my mainframe days that I would love
to scan and convert to text files. (Even better would be to find
a 9-track drive that could read my tapes, but that might be harder.)
I've tried scanning a small listing to PDF, and then played with
ocrmypdf and gocr, but got mostly garbage out. Does anyone have
experience with this sort of thing who might be able to give me
some pointers?

aTdHvAaNnKcSe...

--
/~\ Charlie Gibbs | Microsoft is a dictatorship.
\ / <cgibbs@kltpzyxm.invalid> | Apple is a cult.
X I'm really at ac.dekanfrus | Linux is anarchy.
/ \ if you read it the right way. | Pick your poison.

Charlie Gibbs <cgibbs@kltpzyxm.invalid> wrote:
> I have a lot of listings from my mainframe days that I would love
> to scan and convert to text files. (Even better would be to find
> a 9-track drive that could read my tapes, but that might be harder.)
> I've tried scanning a small listing to PDF, and then played with
> ocrmypdf and gocr, but got mostly garbage out. Does anyone have
> experience with this sort of thing who might be able to give me
> some pointers?

1) Try Tesseract -- https://en.wikipedia.org/wiki/Tesseract_(software)
2) Scan at a fairly high res (300 or 600 dpi)

Sometimes thresholding to 2-color black and white also works well --
but you'll have to test on samples to see.

The wikipedia page for Tesseract has some more suggestions:

Tesseract's output will have very poor quality if the input images
are not preprocessed to suit it: Images (especially screenshots)
must be scaled up such that the text x-height is at least 20
pixels,[13] any rotation or skew must be corrected or no text will
be recognized, low-frequency changes in brightness must be high-pass
filtered, or Tesseract's binarization stage will destroy much of the
page, and dark borders must be manually removed, or they will be
misinterpreted as characters.[14]

In comp.os.linux.misc, Charlie Gibbs <cgibbs@kltpzyxm.invalid> wrote:
> I have a lot of listings from my mainframe days that I would love to
> scan and convert to text files. (Even better would be to find a
> 9-track drive that could read my tapes, but that might be harder.)
> I've tried scanning a small listing to PDF, and then played with
> ocrmypdf and gocr, but got mostly garbage out. Does anyone have
> experience with this sort of thing who might be able to give me some
> pointers?

AEK, who has http://www.bitsavers.org/, does a ton of old document
scanning and describes his tooling on his website. He uses Acrobat for
the OCR step, but describes it as the most time consuming part.

Check the archive.org blog, too, in the past they used to describe their
methods and probably continue to do so.

AEK may know where you can get old tapes converted, he has some
connection with a computer history museum.

Elijah
------
the issues with OCR make one despair for real AI vision

On 2022-02-23 13:54, Charlie Gibbs wrote:
> I have a lot of listings from my mainframe days that I would love
> to scan and convert to text files. (Even better would be to find
> a 9-track drive that could read my tapes, but that might be harder.)
> I've tried scanning a small listing to PDF, and then played with
> ocrmypdf and gocr, but got mostly garbage out. Does anyone have
> experience with this sort of thing who might be able to give me
> some pointers?

Try asking on the cctalk mailing list from http://www.classiccmp.org/

There have been recent discussions on the list (check the archives on
their website) about doing OCR on old code listings. There are also
people on that list with working 9-track tape drives that could either
read the tapes for you or help you set up a drive of your own to do it
yourself.

Charlie Gibbs <cgibbs@kltpzyxm.invalid> wrote:
> I have a lot of listings from my mainframe days that I would love
> to scan and convert to text files. (Even better would be to find
> a 9-track drive that could read my tapes, but that might be harder.)
> I've tried scanning a small listing to PDF, and then played with
> ocrmypdf and gocr, but got mostly garbage out. Does anyone have
> experience with this sort of thing who might be able to give me
> some pointers?

If your plan is to share the scans online, then one option is to
simply upload them to the Internet Archive, who automatically OCR
uploaded documents. If you upload a PDF they will convert that
into a new OCRed PDF, as well as plain text. Or you can also upload
a .zip full of numbered images (001.png, 002.png, 003.png...) and
they'll convert that as well (they had instructions somewhere with
the exact details of how to upload such scans).

Their processing seems to be very good, and may be better than can
be achieved with free software. Obviously you'll still need to make
sure that the resolution is high enough (I usually use 300dpi), and
that the pages are completely flat against the scanner (if
possible).

--
__ __
#_ < |\| |< _#

On 2022-02-23, Eli the Bearded <*@eli.users.panix.com> wrote:

> In comp.os.linux.misc, Charlie Gibbs <cgibbs@kltpzyxm.invalid> wrote:
>
>> I have a lot of listings from my mainframe days that I would love to
>> scan and convert to text files. (Even better would be to find a
>> 9-track drive that could read my tapes, but that might be harder.)
>> I've tried scanning a small listing to PDF, and then played with
>> ocrmypdf and gocr, but got mostly garbage out. Does anyone have
>> experience with this sort of thing who might be able to give me some
>> pointers?
>
> AEK, who has http://www.bitsavers.org/, does a ton of old document
> scanning and describes his tooling on his website. He uses Acrobat for
> the OCR step, but describes it as the most time consuming part.

I've scanned a lot of old mainframe manuals and uploaded them to Bitsavers.
AEK seems to know a lot about PDFs - he compressed my uploads to a third
their size without noticeable loss of quality. Also, he apparently OCRed
them because the reprocessed files have searchable text; pdftotext can
extract it.

> Check the archive.org blog, too, in the past they used to describe their
> methods and probably continue to do so.
>
> AEK may know where you can get old tapes converted, he has some
> connection with a computer history museum.

Thanks, I'll look into that.

--
/~\ Charlie Gibbs | Microsoft is a dictatorship.
\ / <cgibbs@kltpzyxm.invalid> | Apple is a cult.
X I'm really at ac.dekanfrus | Linux is anarchy.
/ \ if you read it the right way. | Pick your poison.