Text layer not aligned with original document

[wpzdm]

Describe the issue When selecting a word, last character is always out of selection. The problem should be clear from the following snapshot: https://imgur.com/f12zGYc This problem appears on all files I tried.

Example file https://gofile.io/?c=ptsFtL

• [x] This is the input file
• [x] The file contains no personal or confidential information
• [ ] I am the copyright holder for this file
• [ ] I permit this file to be included in the OCRmyPDF test suite under the CC-BY-SA 4.0 license
• [ ] I am not the copyright holder, but this file is available under a free software license

System:

• OS: macOS 10.14.5
• OCRmyPDF Version: 9.0.3

[jbarlow83]

The imgur link didn't work unfortunately.

Please upload a PDF. It's very unlikely I will be able to investigate this issue without an example input PDF.

[wpzdm]

I uploaded an input PDF. You can use --force-ocr to override the original text layer. Note that the problem is unrelated to force commend.

BTW, maybe I did not make it clear enough, the problem is in the display of selection area, and the snapshot was meant to show it (not to use as an input file). Here, I selected the word 'quantum', but as you can see from the snapshot, the character 'm' was not selected (but I can still copy the whole word with this selection).

[jbarlow83]

The alternative renderer --pdf-renderer hocr does a better job here. Note that it has issues with non-Latin text.

I believe this is a regression in Tesseract 4.1.0.

Related past issue: https://github.com/tesseract-ocr/tesseract/issues/1900

[wpzdm]

Thank you! hocr did well in this issue, but it seems worse than default renderer in detecting and separating words, particularly in the content tables.

Here is an example: https://gofile.io/?c=ZE8RQG. Compare page 13 of PDF file, hocr detected none of the chapter titles, and did not separate some words like 'Maximal inequality' and 'Kolmogorov Existence Theorem'. BTW, both renderers failed to separate all the occurrences of the phrase 'Section summary'.

Maybe I should open another issue?

[jbarlow83]

Inserting spaces between words is the job of the PDF viewer due to unfortunate design decisions in the early days of PDF. macOS Preview also does a particular poor job of this.

[wpzdm]

Thank you. I use Clearview on Mac. I don't know if it shares the same underlying with Preview.

[jbarlow83]

too old

[Tarek-Hasan]

There's a fork of Tesseract, which fixed text alignment problem in Tesseract here.