Hi everyone,
We are experiencing an issue with OCR on Arabic PDF documents in both Laserfiche Cloud and Laserfiche 12 Self-Hosted.
The extracted text does not match the original Arabic text in the PDF. It also appears that the Arabic text is being extracted in reverse. Since Arabic is read from right to left, the OCR output seems to be in the opposite direction, causing the words and characters to be in the wrong order.
To test this, we performed the following:
-
Laserfiche Cloud: After generating searchable text, we copied a short Arabic word directly from the PDF and performed a Document Text Search. No results were found.
-
Laserfiche 12 Self-Hosted: We performed the same test. After generating searchable text, we copied the same short Arabic word from the PDF and performed a Document Text Search. Again, no results were found.
We also tried the following in Laserfiche 12 Self-Hosted:
-
Changed Tools > Options > Generate Text > Language from English to Arabic.
-
Tried both enabling and disabling Decolumnize Text.
However, the issue remained the same.
From these tests, we found that the Arabic text copied from the original PDF does not match the OCR-generated text stored by Laserfiche.
However, if we copy the Arabic text from the OCR-extracted text and use it in Document Text Search, the document is found. This indicates that the OCR-generated text is different from the original Arabic text in the PDF, which may cause search accuracy issues.
Summary:
-
The OCR-extracted Arabic text does not match the original Arabic text in the PDF/image.
-
Searching using the original Arabic text from the PDF returns no results.
-
Searching using the OCR-extracted Arabic text returns the document, but this is an accuracy issue because the extracted text is not the same as the original document.
Could you please advise:
-
Are there any settings that should be configured for Arabic OCR?
-
Is this a known issue with Arabic OCR or right-to-left text recognition?
Please see the attached screenshot for reference.