Hi,

It seems that Tika indexes PDFs from LTR instead of RTL when indexing Arabic/Hebrew PDFs. I have read about this issue herehttp://stackoverflow.com/questions/7834401/solr-for-arabic and it seems that the solution lies in including ICU4J in the project.... But I have no idea on how this can be done. Can anyone help with this?

Other than the issue mentioned above I have found the following reference to the same problem. The solution according to http://www.linnovate.net/blog/apache-solr-search-hebrew-and-probably-ara... lies in making PDFbox handle the PDFs instead of Tika. I find this answer a bit weird since according to my knowledge Tika itself uses PDFbox.

Your input on this would be very much appreciated.

Comments

bijumon.ta’s picture

bijumon.ta commented

Dear Team,

I am also having the same issue. Arabic PDF documents not indexing from RTL. It is indexed as LTR language mode. I tried indexing with multilingual too but not indexing as expected to RTL languages.I think this was not issue with tika, because it already uses PDF box.

please run code this in command line. It will show the exact data as in the PDF.

java '-Dfile.encoding=UTF8' -cp '/opt/tika-1.0' -jar '/opt/tika-1.0/tika-app-1.2.jar' -t '/var/www/htdocs/MySite/sites/all/modules/apachesolr_attachments/tests/arabic_ar.pdf'.

So this was not the issue with tika, it was the issue in this module. When i checked the database , it stores as LTR format. May be that will be the issue with this module. Please give us a solution to resolve this.

Regards.
Bijumon.T.A

mmff’s picture

mmff commented

Hi,
I encountered the same problem:
Tika indexes Arabic PDF from Leblem with .docx arabic file.
So, if any one have any suggestion, it will be very appreciated.
best regards

mmff’s picture

mmff commented
Issue summary: View changes

grammatical mistake