Sanskrit OCR

[John.Powers at anu.edu.au]

David Dargie asks:

>I think I am safe in assuming that no-one has yet developed an optical
>character recognition (OCR) system for the devanagari script.

In the case of a typeset manuscript (as opposed to a handwritten one), it's
theoretically possible, but the time involved to make it work is probably
prohibitive, since the program would have to be taught how to read
different manuscripts. With the new teachable programs, typeset manuscripts
could be read with varying degrees of accuracy, but in the case of
consonant clusters, there would inevitably be some errors, and it would
probably be much faster to type it.

go, but I have not heard any more about it.
>
>In any case, does anyone know of an OCR system that is accurate for roman
>transliteration.  Has anyone tried to input text in this way?
>
I have done this, but the problem is that I haven't seen any program that
can accurately read diacritical marks. Some of them do come out pretty
well, and can be changed globally, but there are so many errors that it's
still faster to type it manually, and probably generally more accurate.

John Powers
Faculty of Asian Studies
Australian National University