Extract HOCR from searchable PDF

[thwfqecj]

Thank you so much with your great works!

But I wonder if it is possible to extract HOCR from searchable PDF, I mean, PDFs that are already combined with HOCR, I haven't find any tools to do that for me...

[stweil]

Nor do I know such tools. The tool pdftohtml can extract XML from PDF, and there is an issue for ocr-fileformats to convert that XML to hOCR, so that combined tools would do the job.

[jsbien]

Quote/Cytat - Stefan Weil <notifications@github.com> (wto, 1 sie 2017, 15:49:41):

Nor do I know such tools. The tool `pdftohtml` can extract XML from PDF, and there is an [issue for ocr-fileformats](UB-Mannheim/ocr-fileformat#57) to convert that XML to hOCR, so that combined tools would do the job.

Or you can convert PDF to DjVu and export hOCR from DjVu with Jakub Wilk's tools https://jwilk.net/software/pdf2djvu https://jwilk.net/software/ocrodjvu Regards Janusz

…

-- Prof. dr hab. Janusz S. Bień - Uniwersytet Warszawski (Katedra Lingwistyki Formalnej) Prof. Janusz S. Bień - University of Warsaw (Formal Linguistics Department) jsbien@uw.edu.pl, jsbien@mimuw.edu.pl, http://fleksem.klf.uw.edu.pl/~jsbien/

[thwfqecj]

@stweil @jsbien

Thanks for your comment!
I'm trying to use pdftohtml. Actually I want to make my search-able pdf slimmer. It seems I have to change all the pages to single html files and then back to pdf...But it works!

As for ocrodjvu,

ocrodjvu is a wrapper for OCR systems, that allows you to perform OCR on DjVu files.

It looks like a OCR tool, not fitting my needs. But Thank you.

[jsbien]

ocrodjvu is distributed with djvu2hocr which is what you may need,

[giancarlobi]

Many thanks @jsbien pdf2djvu + djvu2hocr works like a charm !!

[thwfqecj]

@jsbien Thank you for your kindly reply! Sorry for I didn't see ocrodjvu on its official site...Guess it should works!

[jsbien]

Quote/Cytat - thwfqecj <notifications@github.com> (Wed 18 Oct 2017 10:58:13 AM CEST):

@jsbien Thank you for your kindly reply! Sorry for I didn't see ocrodjvu on its official site...Guess it should works!

What do you mean by its official site? For me it is https://jwilk.net/software/ocrodjvu

[JensHumrich]

Hey,
I stumbled upon this old thread. I can confirm that the solution works...

pdf2djvu -o test.djvu test.pdf
python2 /mnt/mem/temp/ocrodjvu/ocrodjvu test.djvu -o ocrfile
python2 /mnt/mem/temp/ocrodjvu/djvu2hocr ocrfile > output.hocr

[jsbien]

For a searchable PDF the second step should be skipped, otherwise instead of the original text you get the result of OCR.

[JensHumrich]

Wow. Thanks a lot. This is really an important information.

[mattdeeperinsights]

I would recommend using Python package pdftotree to get the hocr automatically, it's so easy.

Get requirements:

• Get latest Java (8+) if you don't already have it
• Get latest ImageMagick

Pip the package: pip3 install pdftotree and then it's as simple as this:

import pdftotree
hocr_result = pdftotree.parse('path/to/your.pdf')

Enjoy.

[rmast]

Hey, I stumbled upon this old thread. I can confirm that the solution works...

pdf2djvu -o test.djvu test.pdf
python2 /mnt/mem/temp/ocrodjvu/ocrodjvu test.djvu -o ocrfile
python2 /mnt/mem/temp/ocrodjvu/djvu2hocr ocrfile > output.hocr

I can now say it doesn't work for either a PDF or a DjVu with searchable text coming from GScan2PDF.
Te resulting ocrfile is still big and containing pages that are mentioned by djvu2hocr, however the resulting output.hocr contains nothing. The contents of the file look like a djvu, and renamed to djvu are viewable by a djvu-viewer. They show no hidden text.

To get the HOCR from the searchable DjVu just apply djvu2hocr on the djvu and skip ocrodjvu