Allow to specify PAPERLESS_OCR_MODE and PAPERLESS_OCR_LANGUAGE when doing "redo OCR"

[timaschew]

My default PAPERLESS_OCR_MODE is skip_noarchive and my default language is "deu" (German) and in most cases it's fine.

But somemtimes:

1. I upload a PDF which was already scanned with bad OCR (cyrillic chars in a German PDF)
2. I upload a document in another language (sometimes, English, sometimes russian, etc.)

To fix use case 1 I would need to pass redo as PAPERLESS_OCR_MODE. Because I can't do this, I need to shutdown the server, change the PAPERLESS_OCR_MODE to redo, then press in the UI "redo OCR" and then change PAPERLESS_OCR_MODE back to my default.

It's similar to use case 2 where I need to change the language just for a single document where I want to redo the OCR.

[exxus]

I have a very similar use case.
My scan software automatically runs ocr. But sometimes the file is not rotated correctly. Than the consumed pdf contains wrong text data.
In that case the new redo button (#1139, #1277) does nothing, since by default pages already containing text are ignored.
In my opinion, the redo ocr button should always pass the "--redo-ocr" option to ocrmypdf, since I manually force an ocr and even get a warning about losing the old text.

I agree that changing the conf or env files each time is very tedious.

For use case 2, you could pass multiple languages e.g., I pass PAPERLESS_OCR_LANGUAGE=deu+eng since I have both English and German files. It than selects the correct language automatically.

Maybe it is possible to add the ocr options to the UI settings? But there are many options and it may clutter the UI too much.

[devtonhere]

Alternatively the behaviour of the UI redo button can be changed to set PAPERLESS_OCR_MODE=redo temporary/just for that scan.

And on the settings page a new input field could be provided to overwrite PAPERLESS_OCR_LANGUAGE.

I guess this would require less UI changes and would also solve the problem to restart the server to do a "real redo" or change the language.

[flesser]

05.09.2022 23:07:22 ***@***.***: Hiermit erhalten Sie Ihre Rechnung Rechnung-715966EW.pdf Mit freundlichen Grüßen Elektrowelt Zwickau GmbH Impressum: Elektrowelt Zwickau GmbH Herschelstraße 9 08060 Zwickau Tel.: 0375 / 27 04 99 50 Fax: 0375 / 27 04 99 69 E-Mail: ***@***.***

[furkanusta]

It would be nice to have this in the UI but in the meantime you can do it in the terminal with docker exec

docker exec -d  -e "PAPERLESS_OCR_MODE=force" paperless_webserver_1 document_archiver --overwrite --document 15

If this is a common occurence you can have a small bash function which takes environment variables or document id as parameter.

[timaschew]

Thanks, that is a good workaround meantime :)

[baxerus]

I would also "vote" for having at least having the "Redo OCR" button really always have the OCR mode set to force.
I mean pressing a button labeled like this is already a very explicit expression of the users will to redo the OCR on a file not matter what 😉

I also would really love to see that pressing the "Redo OCR" button also creates a "File Task". Yes, you could check the log to see if something happened, but having a task in the tasklist (that then maybe would also show if something was wrong) would be so much more the right thing. But this in itself is probably it's own feature request

[stepanov1975]

+1 to change "Redo OCR" behavior to always force

[exhuma]

I'm currently evaluating paperless-ngx and setting the OCR language is very important. I'm from Luxembourg. We are plurilingual and - due to the size of the country - have a lot of interactions with foreign companies. Even for personal use-cases. It is common here to find the best deal not only between companies, but also tapping into our neighbouring countries. We also have many foreign companies working here.

So it is commonplace to receive letters in three languages (French, German and English) in a fairly even split.

By using only one language in Paperless NGX I can only benefit from OCR in about a third of the scanned documents.

I would expect a similar situation to exist in border-regions of in other countries as well. Luxembourg just happens to be one big border-region 😆

[stumpylog]

I think your situation is already handled by PAPERLESS_OCR_LANGUAGE. It can be a combination of languages, so setting it to the 3 mentioned above, in combination with PAPERLESS_OCR_LANGUAGES to install them if needed, will help recognize those in scanned documents. The value would look something like eng+fra+deu

For digital documents, the existing text is used as much as possible, making the language setting less important.

[exhuma]

Ah indeed. This is even noted in the bash script prompts. I missed this in the docs. Thanks for taking the time to reply. I apreciate it.

[almereyda]

There are related discussions in:

• Try importing a file again without OCR if OCR fails #3032
• [Feature Request] Reprocess documents with different settings #3090
• [Feature Request] Redo OCR with specific language #4234

[architomsn]

+1 to change "Redo OCR" behavior in the ui for specific documents

[deajan]

+1, "Redo OCR" should relaunch OCR process, regardless of settings. Makes totally sense.

[rvdv01]

Agreed, I also had some PDFs which already had text included, but that text was 'scanned with a trial version of program xxxx' instead of the actual text in the PDF :-)
So a real force redo button would really help when I get such documents.