Are there any OCR programs that can process Syriac?

March 14, 2008 8:16 PM

Are there any OCR programs that can read Syriac?

This one's a pretty straightforward one, I'm just looking for an OCR program that can recognize Syriac and output it into a digital document. Anyone have any leads on software that may do this?

posted by perpetualstroll to religion & philosophy (3 comments total)

Considering what it looks like, and how vitally important the distinction between where one letter starts and another ends in OCR is, I'd say it's a significantly harder problem than OCRing cursive English.

That said, Googling leads me to this: William Clocksin and Prem Fernando are the experts, this is the relevant paper, and Clocksin describes having a program on his page linked above.

I suggest emailing him and asking if his program is available. Whether he'll say "sure, here's a copy" or "I rent that out at $5000 per hour!" is uncertain, but it's worth asking.
posted by aeschenkarnos at 8:46 PM on March 14

Further interesting stuff: the 242-page PDF PhD thesis of Mohammed S. M. Khorsheed, "Automatic Recognition Of Words In Arabic Manuscripts", a student of William Clocksin's, in June 2000.

The open source program Tesseract, and a quick perusal of its associated files indicates that the most typographically complex language it can handle so far is Vietnamese, which is a Roman script variation with 29 base letters and 6 tonal indicators.

I think the short answer is it doesn't really exist, yet. But I'm sure there's a fortune to be earned for making it.
posted by aeschenkarnos at 9:03 PM on March 14

GOCR can be trained to recognise new alphabets, but would have trouble with being able to separate letters perhaps.

I suggest posting to the GOCR mailing list (on jocr.sf.net), and linking to some high res images of some Syriac text, you may get some interest.
posted by Jerub at 3:56 AM on March 15