Best way to handle Arabic OCR - what results have you had? what software to use?

(4 posts) (2 voices)

Tags:

Elli Mylonas
Member
Does anyone have experience, positive or otherwise, with Arabic OCR? How did you go about it, what did you use and did it require special configuration.
Tweet this question
Posted 5 years ago Permalink
Trip Kirkpatrick
Member

I haven't used it for this case, but I had some minor POC success with Tesseract. It was non-trivial to install for me, but YMMV. Nicely, they have added Arabic as a covered script in the most recent version. Out of curiosity, is it handwriting or typeset?

Posted 5 years ago Permalink
Elli Mylonas
Member

Replying to @Trip Kirkpatrick's post:

typeset very consistently in the late '50s, but the letters look a little "fancy" and have a lot of very thin lines. I don't read Arabic, so can't describe in more detail.

Posted 5 years ago Permalink
Trip Kirkpatrick
Member

In unrelated research, I found a tesseract training GUI written by a web developer: http://pp19dd.com/2010/09/tesseract-ocr-boxfile-ajax-editor/
Since I haven't had to train tesseract yet, I'm not entirely confident of reading that page well, but it looks like it might be useful if you find the existing tesseract Arabic data lacking.

It was originally written in 2010, but has been updated for tesseract v3.

Posted 5 years ago Permalink

Reply

You must log in to post.