[
https://issues.apache.org/jira/browse/TIKA-4231?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17832291#comment-17832291
]
Tilman Hausherr commented on TIKA-4231:
---------------------------------------
I have attached an extraction with pdfbox 2.0.31: [^arabic-pdfbox.txt]
is this better, or not? I've added a BOM and removed the 00 bytes. In the tika
extraction there are many "ef bf bd" bytes instead which is the utf8
replacement character �.
A possible explanation why Adobe Reader works better is that this file uses the
"ActualText"-feature which PDFBox doesn't support (PDFBOX-3248).
> Parsing Arabic PDF is returning bad data
> ----------------------------------------
>
> Key: TIKA-4231
> URL: https://issues.apache.org/jira/browse/TIKA-4231
> Project: Tika
> Issue Type: Bug
> Affects Versions: 2.6.0, 2.9.1
> Environment: I am using Java 18. And using maven dependency
> tika-parsers-standard-package
> ([https://mvnrepository.com/artifact/org.apache.tika/tika-parsers/2.6.0)]
>
> Reporter: Aamir
> Priority: Major
> Attachments: arabic-pdfbox.txt, arabic.pdf, arabic.txt
>
>
> Attached is a PDF with arabic text in it.
> When parsed using tika version 2.6.0 or 2.9.1, it produces gibberish
> characters.
> The generated text doc is also attached which contains the parsed text.
> Most of the other Arabic PDFs parse fine, but this one is giving this output.
--
This message was sent by Atlassian Jira
(v8.20.10#820010)
