[INDOLOGY] OCR
Rolf Heinrich Koch
rolfheiner.koch at gmail.com
Sun May 11 14:00:10 UTC 2025
Yes, Patrick, this is possible- including some errors. I give you an
example:
1. I loaded up this image to ChatGPT (a verse from AK):
2. ChatGPT produced this transkription (searchable):
धूमग्निस्तरणिर्मित्रशत्रुभानुगभस्तिरोचनः ।
विभावसुसुप्रीहपतिस्तेपवांस्तपतिर्यपातिः ॥ ३० ॥
3. a transliteration
dhūmagniḥ taraṇir mitraśatrubhānugabhastirocanaḥ |
vibhāvasuḥ suprīhaḥ patiḥ te pavāṁs tapatiḥ ravipātiḥ || 30 ||
4. and this translation:
Dies ist ein Vers aus dem *Amarakosha*, einem bekannten
Sanskrit-Thesaurus. Der Vers listet *Synonyme für die Sonne (Surya)* auf:
*„Dhūmagni, Taraṇi, Mitra, Śatru, Bhānu, Gabhasti, Rocaṇa,
Vibhāvasu, Suprīha, Pati, Tapavāṁs, Tapatī, Ravipāti“*
→ all dies sind Namen oder Bezeichnungen für die *Sonne*.
ChatGPT do not care about any font, by which the devanagari text was
produced and needs more training as you can see.
Best
Heiner
Am 10.05.2025 um 00:07 schrieb Patrick Olivelle via INDOLOGY:
> Dear Friends:
>
> I am wondering whether with the advance of AI technology we have easy OCR software to read Devanāgarī, easy enough to be used by someone like me!! We have the one prepared by Andrew Ollett, which he generously gave us. But that requires computer knowledge far beyond my reach. Is there on where you can just drop the Devanāgari scan, out pops a searchable file. This is probably a long shot, but I thought I would ask.
>
> With thanks and best wishes,
>
> Patrick Olivelle
>
> _______________________________________________
> INDOLOGY mailing list
> INDOLOGY at list.indology.info
> https://list.indology.info/mailman/listinfo/indology
--
Dr. R. H. Koch - Germany/Sri Lanka
www.rolfheinrichkoch.wordpress.com
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://list.indology.info/pipermail/indology/attachments/20250511/4e14b29d/attachment.htm>
-------------- next part --------------
A non-text attachment was scrubbed...
Name: ouMY9lwiz9ZAnifs.png
Type: image/png
Size: 32075 bytes
Desc: not available
URL: <https://list.indology.info/pipermail/indology/attachments/20250511/4e14b29d/attachment.png>
More information about the INDOLOGY
mailing list