• OCR

    From Wendelin Uez@wuez@online.de to alt.windows7.general on Mon Oct 5 08:14:50 2026
    From Newsgroup: alt.windows7.general

    I have a lot of printed documents to digitze, no graphics, no tables, either running text only or lines with 2 columns like "name <tab> text", but some words in bold, and some headers in bigger character sizes.

    Is there any OCR software available for WINXP, WIN 7, or WIN 8.1 which can recognize this structure and properties when scanning directly or reading a pdf scan (which is probably an embedded picture)?


    --- Synchronet 3.22a-Linux NewsLink 1.2
  • From Arkadiusz 'Black Fox' Artyszuk@asdfadsasdfa@atomicmail.io to alt.windows7.general on Mon Oct 5 10:24:10 2026
    From Newsgroup: alt.windows7.general

    Wendelin Uez wrote:

    I have a lot of printed documents to digitze, no graphics, no tables,
    either running text only or lines with 2 columns like "name-a <tab>
    text", but some words in bold, and some headers in bigger character sizes.

    Is there any OCR software available for WINXP, WIN 7, or WIN 8.1 which
    can recognize this structure and properties when scanning directly or reading a pdf scan (which is probably an embedded picture)?

    Free:
    - an older Tesseract 3.0X Windows build with an older gImageReader
    release compatible with that Tesseract version,
    - FreeOCR

    Commercial:
    - ABBYY FineReader 12
    --
    Regards
    Arkadiusz 'Black Fox' Artyszuk
    --- Synchronet 3.22a-Linux NewsLink 1.2
  • From Java Jive@java@evij.com.invalid to alt.windows7.general on Mon Oct 5 10:44:57 2026
    From Newsgroup: alt.windows7.general

    On 2026-10-05 09:24, Arkadiusz 'Black Fox' Artyszuk wrote:
    Wendelin Uez wrote:

    I have a lot of printed documents to digitze, no graphics, no tables,
    either running text only or lines with 2 columns like "name-a <tab>
    text", but some words in bold, and some headers in bigger character
    sizes.

    Is there any OCR software available for WINXP, WIN 7, or WIN 8.1 which
    can recognize this structure and properties when scanning directly or
    reading a pdf scan (which is probably an embedded picture)?

    Some free PDF readers, for example Tracker Software's PDF-XChange Lite
    (there is also PDF-XChange Editor but I can't remember if that's free)
    can read an existing image only PDF, OCR it, and allow you to save the
    OCR layer with the existing PDF.

    Free:
    - an older Tesseract 3.0X Windows build with an older gImageReader
    release compatible with that Tesseract version,
    - FreeOCR

    Commercial:
    - ABBYY FineReader 12

    ABBYY came bundled with my Epson V600 scanner, 'though I've never tried
    it because I have LibreOffice, so can't remember whether you have to pay
    to register and use it.
    --

    Fake news kills!

    I may be contacted via the contact address given on my website: www.macfh.co.uk

    --- Synchronet 3.22a-Linux NewsLink 1.2
  • From Steve Hayes@hayesstw@telkomsa.net to alt.windows7.general on Mon Oct 5 15:18:25 2026
    From Newsgroup: alt.windows7.general

    On Mon, 5 Oct 2026 08:14:50 +0200, "Wendelin Uez" <wuez@online.de>
    wrote:

    I have a lot of printed documents to digitze, no graphics, no tables, either >running text only or lines with 2 columns like "name <tab> text", but some >words in bold, and some headers in bigger character sizes.

    Is there any OCR software available for WINXP, WIN 7, or WIN 8.1 which can >recognize this structure and properties when scanning directly or reading a >pdf scan (which is probably an embedded picture)?

    I think there are some cell phone ocr apps that can tead to text or
    rtf format. Then copy the resulting file from your cellphone to cyour
    computer and edit there with your normal word processing software.
    --
    Steve Hayes from Tshwane, South Africa
    Web: http://www.khanya.org.za/stevesig.htm
    Blog: http://methodius.blogspot.com
    E-mail - see web page, or parse: shayes at dunelm full stop org full stop uk --- Synchronet 3.22a-Linux NewsLink 1.2
  • From Paul@nospam@needed.invalid to alt.windows7.general on Mon Oct 5 14:23:24 2026
    From Newsgroup: alt.windows7.general

    On Mon, 10/5/2026 2:14 AM, Wendelin Uez wrote:
    I have a lot of printed documents to digitze, no graphics, no tables,
    either running text only or lines with 2 columns like "name-a <tab>-a text", but some words in bold, and some headers in bigger character sizes.

    Is there any OCR software available for WINXP, WIN 7, or WIN 8.1 which
    can recognize this structure and properties when scanning directly or reading a pdf scan (which is probably an embedded picture)?

    The details are important.

    I did a conversion of a document with a number of test cases on it.
    I had two columns of text, financial entry, dollar amount. This is a portion
    of the test document. The tesseract test sections are there, as most freewares will have tesseract as the base OCR.

    Accounts Incoming DejaVuSansMono Amount 0 O o test tesseract

    Royalty Bank $ 1.39 \
    $ 10.00 \__ A different font
    $1000.00 Oh oh 0h test tesseract /

    If I wipe over the converted text (sandwiched in the PDF), and paste it, I get this.

    Bold Title16
    Big Bold Title32
    Big Italics Title24
    Accounts Incoming DejaVuSansMono \
    Royalty Bank \
    TrackWinnings DogTrack \
    Taxes Due \___ The glyphless font used, would not actually render
    Amount 0 0 o test tesseract / characters as such, on the screen, in a PDF reader.
    $ / The glyphless font is used to make it possible to
    1.39 / copy and paste the text strings. There is (as usual)
    $ 10.00 / no positional text in the conversion. All my
    $1000.00 Oh oh Oh test tesseract / formatting is gone.

    Tesseract is still failing the test where we check
    capital O, small o, and the number zero. It cannot
    reliably recognize them, unless there is good language context
    of some sort.

    Notice the columnar relationship is destroyed.

    I'm not even sure pdftotext can resurrect this, because
    the converted text was not spaced to capture the position
    on the page in the first place. There was a version of that,
    before some putz rewrote it, where one of the properties
    was, it kept text on the same baseline, as text on the
    same line in the txt output file. Which is not true of the
    version today. The old pdftotext, could in effect "re-glue"
    characters and make a longer string of them.

    Summary: Basically, forget it.
    It's still "too much science project, not practical enough".
    Sure, I have captured "some text at a 97% accuracy, losing format information".
    But, what's the point. No user actually wants my abysmal result.

    *******

    Your best bet is to find a tool that does "pixmap to Microsoft Word".
    It still may not capture positional information properly, but it is
    a different avenue to explore.

    One of the Qwen LLM-AI models, may be intended to do this
    sort of work, but the "user manual" discouraged me from even trying.
    It's not often a user manual manages to do that.

    In a Google search, I don't always get an AI Overview, but I got one this time.

    AI Overview

    To convert a scanned PDF to an editable Word document, you must use OCR
    (Optical Character Recognition) because a scanned PDF is just a picture
    of text rather than a real text layer. You can use online tools or desktop
    software to complete this task.

    Method 1: Use Online OCR Converters (Fastest) <=== debate-able

    Go to a free online tool that supports OCR, such as the Adobe
    Acrobat PDF to Word Tool or Smallpdf PDF to Word.
    Upload your scanned PDF file from your computer or cloud storage.
    Turn on or enable the OCR option if prompted.
    Click Convert and download your new editable file. <=== online tools can have usage
    limits or they can entirely
    fail after the upload.

    Method 2: Use Microsoft Word (Desktop)

    Open Microsoft Word on your computer.
    Click File > Open and select your scanned PDF file.
    Click OK on the warning message explaining that Word will convert the PDF
    into an editable document using background OCR.

    Review the text, fix any formatting shifts, <=== really ? pay money, this happens ?
    and save the file as a document.

    If you'd like, let me know: Is your scan faint, blurry, or very long? \
    Do you have complex tables or images you need to preserve? \_ Hahaha
    I can recommend the best specific tool or cleanup tips for your document type. /

    This is what computers are for, to do half a job.

    I'd tell you about my adventure getting the AI to help
    me remove the background image from my converted text document,
    at first I was amazed at the Python code it wrote, but when
    the code didn't work worth a darn, even some clouds formed
    outside my window, in support.

    It's probably cheaper to hire a secretary, to type it up by hand :-)

    Paul


    --- Synchronet 3.22a-Linux NewsLink 1.2
  • From J. P. Gilliver@G6JPG@255soft.uk to alt.windows7.general,alt.windows7.general on Mon Oct 5 23:21:45 2026
    From Newsgroup: alt.windows7.general

    On 2026/10/5 7:14:50, Wendelin Uez wrote:
    I have a lot of printed documents to digitze, no graphics, no tables, either running text only or lines with 2 columns like "name <tab> text", but some words in bold, and some headers in bigger character sizes.

    Is there any OCR software available for WINXP, WIN 7, or WIN 8.1 which can recognize this structure and properties when scanning directly or reading a pdf scan (which is probably an embedded picture)?


    Windows 7 'group added :-)

    When you say "recognize the structure and properties", do you mean
    actually produce a result file that's in columns, or just not be
    confused by such a layout?
    --
    J. P. Gilliver. UMRA: 1960/<1985 MB++G()ALIS-Ch++(p)Ar++T+H+Sh0!:`)DNAf

    As my chronological age races on, my OldTimers Disease now includes
    Toelio, Kneesles and SmallCox. - Chris Schram in e-s.talk, 2026-6-27
    --- Synchronet 3.22a-Linux NewsLink 1.2
  • From Daniel70@daniel47@nomail.afraid.org to alt.windows7.general on Tue Oct 6 20:03:59 2026
    From Newsgroup: alt.windows7.general

    On 6/10/2026 9:21 am, J. P. Gilliver wrote:
    On 2026/10/5 7:14:50, Wendelin Uez wrote:
    I have a lot of printed documents to digitze, no graphics, no
    tables, either running text only or lines with 2 columns like "name
    <tab> text", but some words in bold, and some headers in bigger
    character sizes.

    Is there any OCR software available for WINXP, WIN 7, or WIN 8.1
    which can recognize this structure and properties when scanning
    directly or reading a pdf scan (which is probably an embedded
    picture)?


    Windows 7 'group added :-)

    Windows 7 'group removed :-) because it was already being sent to the
    windows7 newsgroup, J.P.!!

    When you say "recognize the structure and properties", do you mean
    actually produce a result file that's in columns, or just not be
    confused by such a layout?

    --
    Daniel70
    --- Synchronet 3.22a-Linux NewsLink 1.2
  • From Paul@nospam@needed.invalid to alt.windows7.general on Tue Oct 6 06:11:39 2026
    From Newsgroup: alt.windows7.general

    On Tue, 10/6/2026 5:03 AM, Daniel70 wrote:
    On 6/10/2026 9:21 am, J. P. Gilliver wrote:
    On 2026/10/5 7:14:50, Wendelin Uez wrote:
    I have a lot of printed documents to digitze, no graphics, no
    tables, either running text only or lines with 2 columns like "name
    <tab>-a text", but some words in bold, and some headers in bigger
    character sizes.

    Is there any OCR software available for WINXP, WIN 7, or WIN 8.1
    which can recognize this structure and properties when scanning
    directly or reading a pdf scan (which is probably an embedded
    picture)?


    Windows 7 'group added :-)

    Windows 7 'group removed :-) because it was already being sent to the windows7 newsgroup, J.P.!!

    I think that was intended to be a cross to the Windows 8 group.
    Using your eagle eye, you should have got out your
    spirit level and touched that up :-)

    Paul

    --- Synchronet 3.22a-Linux NewsLink 1.2
  • From Frank Slootweg@this@ddress.is.invalid to alt.windows7.general on Tue Oct 6 10:27:42 2026
    From Newsgroup: alt.windows7.general

    Java Jive <java@evij.com.invalid> wrote:
    [...]
    - ABBYY FineReader 12

    ABBYY came bundled with my Epson V600 scanner, 'though I've never tried
    it because I have LibreOffice, so can't remember whether you have to pay
    to register and use it.

    It's free to use, if you (generic you) got it with your scanner. So
    it's not free to distribute and let someone else use it. (I got it with
    my EPSON PERFECTION V30 scanner.)
    --- Synchronet 3.22a-Linux NewsLink 1.2
  • From Daniel70@daniel47@nomail.afraid.org to alt.windows7.general on Tue Oct 6 21:34:37 2026
    From Newsgroup: alt.windows7.general

    On 6/10/2026 9:11 pm, Paul wrote:
    On Tue, 10/6/2026 5:03 AM, Daniel70 wrote:
    On 6/10/2026 9:21 am, J. P. Gilliver wrote:
    On 2026/10/5 7:14:50, Wendelin Uez wrote:
    I have a lot of printed documents to digitze, no graphics, no
    tables, either running text only or lines with 2 columns like
    "name <tab> text", but some words in bold, and some headers in
    bigger character sizes.

    Is there any OCR software available for WINXP, WIN 7, or WIN
    8.1 which can recognize this structure and properties when
    scanning directly or reading a pdf scan (which is probably an
    embedded picture)?


    Windows 7 'group added :-)

    Windows 7 'group removed :-) because it was already being sent to
    the windows7 newsgroup, J.P.!!

    I think that was intended to be a cross to the Windows 8 group. Using
    your eagle eye, you should have got out your spirit level and touched
    that up :-)

    Paul

    Now that I look back and see the mention of WIN 8.1, you might be right,
    Paul, but my mind-reading ability is failing me. ;-)
    --
    Daniel70
    --- Synchronet 3.22a-Linux NewsLink 1.2
  • From Dr Newsgroup@invalid@invalid.invalid to alt.windows7.general on Tue Oct 6 16:28:48 2026
    From Newsgroup: alt.windows7.general

    On 06/10/2026 11:34, Daniel70 wrote:
    but my mind-reading ability is failing me.

    Probably you need cataract surgery? It's a 10 minute procedure if you
    qualify for this! and your health is otherwise good enough.




    --- Synchronet 3.22a-Linux NewsLink 1.2
  • From Wendelin Uez@wuez@online.de to alt.windows7.general,alt.windows7.general on Fri Oct 9 18:53:35 2026
    From Newsgroup: alt.windows7.general

    When you say "recognize the structure and properties", do you mean
    actually produce a result file that's in columns, or just not be
    confused by such a layout?

    The documents I have to convert contain to types of text layozt, either floating text with differences in font size and use of bold (which should be represented in the target document), or a tabbed list containig per line a short text with some words only, a tab space, and a second text group.
    Nothing else, but both formats mixed in the same document.

    --- Synchronet 3.22a-Linux NewsLink 1.2
  • From Paul@nospam@needed.invalid to alt.windows7.general on Tue Oct 13 01:59:09 2026
    From Newsgroup: alt.windows7.general

    On Fri, 10/9/2026 12:53 PM, Wendelin Uez wrote:
    When you say "recognize the structure and properties", do you mean
    actually produce a result file that's in columns, or just not be
    confused by such a layout?

    The documents I have to convert contain to types of text layozt, either floating text with differences in font size and use of bold (which should
    be represented in the target document), or a tabbed list containig per line
    a short text with some words only, a tab space, and a second text group. Nothing else, but both formats mixed in the same document.

    A search string might be:

    linux ocr document reconstruction

    The property you are asking for, is document reconstruction,
    where the bold and italics are preserved. Some OCR tools have
    HTML output or "markdown text" format. In Markdown, asterisks
    are used as string modifiers. *text* **text** ***text***
    Those indicate Bold/Italics.

    While an HTML format and a Markdown format can potentially
    represent Bold and Italics, it is not really all that likely
    that the OCR people will attempt to provide that information.

    I looked at LLM-AI OCR items, and most of my efforts were too
    pathetic to succeed. I even tried using a regular LLM-AI,
    pasted an image right into the query box, and... it did OCR!
    But the output text, the stupid tool did not output "verbatim"
    what the OCR revealed. The stupid AI had to "re-write" the text
    and mess it up. Giant fail, as a document reconstruction attempt.
    But it was still pretty funny, to see an LLM-AI not listing OCR
    capability, could actually do OCR when asked. When I used the
    File-Open dialog on the LLM-AI local model hosting software,
    the LLM-AI kept telling me it "could not access the file".
    It was when I pasted a 300DPI image I'd copied, into the
    dialog box, it started to chew on that and after a few minutes,
    it output a (sorta) text representation. I love surprises.

    *******

    Today, I went back to conventional techniques. The host OS was
    Ubuntu 26.04 (as I wanted up-to-date packages and this was an
    attempt to reduce the fuss while trying to get it to work).

    These are examples of what worked in Ubuntu 26.04 .

    $ snap install simple-text-extractor
    simple-text-extractor 1.6 from Gaetan Sencie (stens66) installed

    $ simple-text-extractor # GUI-only, select testdoc-pixmap.pdf inside the tool
    # Don't click any tick boxes.
    # Should make testdoc-pixmap_ocr.pdf as output

    $ pdftotext -layout testdoc-pixmap_ocr.pdf # Makes testdoc-pixmap_ocr.txt

    $ cat testdoc-pixmap_ocr.txt # Dump the text to the screen for a look
    # A good quantity of white space is removed
    # but not all of it is removed.
    # The "Sheet1" and "Page1" labels are from
    # LibreOffice Calc (=Excel).
    Sheet1



    Institution Amount

    JolityFarm $10.00
    PieShop $100.00
    Rebate $1,000.00

    The quick brown fox jumped over the lazy hen.
    Four and twenty black birds all in a row.
    This would be a series of paragraphs if it was
    a series of paragraphs.




    Page 1

    # It has mostly preserved the spacing but removes
    # a lot of the more wasteful white space. This was a Letter format page.
    # The margins were removed.

    ******* Now, we test a second OCR package, one with "unpaper" for deskew and thresholding

    # Advantage of this method, is it is command-line based. You can process
    # more documents in one run this way. The other method was GUI based.

    $ sudo apt install ocrmypdf ocrmypdf-doc # The doc is in HTML format, for Firefox.

    $ ocrmypdf testdoc-pixmap.pdf testdoc-pixmap-ocrmypdf.pdf # No doc cleanup, just convert it.

    $ pdftotext -layout testdoc-pixmap-ocrmypdf.pdf # Makes testdoc-pixmap-ocrmypdf.txt

    $ cat testdoc-pixmap-ocrmypdf.txt # Notice the amount of white space
    # suppressed is slightly different.
    Sheet1



    Institution Amount

    JolityFarm $10.00
    PieShop $100.00
    Rebate $1,000.00

    The quick brown fox jumped over the lazy hen.
    Four and twenty black birds all in a row.
    This would be a series of paragraphs if it was
    a series of paragraphs.




    Page 1

    ******* End-of-experiments *******

    Anyway, that's not remotely "document reconstruction", but
    this is an almost workable alternative. Your "Bold" is not
    preserved. But you might get some usage from this.

    Having a product that can do this on Windows 7 ?
    It is possible, but the software would have to be pretty
    crusty if it was still for sale. Maybe an ABBYY product
    could do it ? Not sure.

    Really, what changed today, is getting "pdftotext" working.
    Once I discovered the -layout option, I was in business.

    Paul
    --- Synchronet 3.22a-Linux NewsLink 1.2