OCR problem in a PDF flie

[mzel]

I have recently installed koreader on a Forma. When trying to lookup a word or reflow a page in a .pdf I get the error message "No OCR results or no language data..." I did install the tesseract data files, but see the same error. Is it possible that the document's language setting affects this? In Book Information Language is N/A. Can I override it somewhere?

[Frenzie]

The document language doesn't really mean anything.

A file like https://github.com/tesseract-ocr/tes...ng.traineddata should go in the /mnt/onboard/.adds/koreader/data folder. If you get that message it probably means you accidentally put it in not quite the right location.

What does the full path to your file look like?

[mzel]

As far as I see it is there:
But what is even stranger the .pdf has the original text layer which is seen by nickel, but not koreader

[mzel]

Ok, I missed two things:
1) A restart was needed after I installed those files
2) I had forced OCR on
This solved the above problems

[mzel]

However all of the above did not get me to the desired end point. I have a magazine in .pdf form with two columns of text and with reflow on I get the single column scaled to the full screen width. Am I expecting too much?

[Frenzie]

Hard to say, I'm not sure what else you'd expect from it? That's kind of its raison d'être.

(That being said, personally I typically prefer fit to content width.)

[mzel]

My idea was to have a free-flow text which adjusts to the width of the screen according to the font size.

[Frenzie]

So your complaint is that the column will only get smaller if you increase the font size but it won't get larger with a smaller font size? I guess that'd be a bug. You could see if you can manage something with k2pdfopt by itself.

[mzel]

My complaint is that the text is not treated as text. The page is treated as a collage of 2-3 images and each of those images is scaled to the width of the page. I'd expect it to treat text as text where the number of characters in a line depends on the width of the screen and font size, the same way it is handled in epub/txt

[Frenzie]

See the attachments for how it behaves for me.

[mzel]

No, I see different behavior. I see two different variants - for small font, and for default and larger (any larger size). For a font larger than 3rd setting from the left it does not scale. See a snapshot of the entire page too

[Frenzie]

In the future, could you please open with that?

You could try playing around with it in k2pdfopt on your PC. Also is this file something that's publicly available?

[viceant]

Excuse me but, in my ereader "Likebook plus ( t80s ) OCR doesn't show any problem, but I can't change text size because it seems not to be enabled - it shows available sizes grey

Edit: sorry, I figured out how to enable it.

[mzel]

Here is the publicly available fragment:
https://litportal.ru/trial/pdf/38836104.pdf
Full file shows even more problems

[Frenzie]

That document seems to be dealt with quite well, except it's a bit of a difficult one in that the number of columns isn't stable. Just setting it to three seems fine though, also where there are only two columns.

Were you by any chance using a nightly instead of the stable? There's a reason the GH bug report template asks for that kind of information. Because if so you need the 2020.04.1 release to prevent some minor cache mismatch issues that were present for a couple of weeks.