[INDOLOGY] Introducing Dharmamitra.org, grammatical analyzers and machine translation for Sanskrit

[Sebastian Nehrdich]

Dear List members,

This is to briefly introduce dharmamitra.org, a project lead by Kurt
Keutzer and myself at BAIR, UC Berkeley, focussing on providing various
GenAI-driven applications for classical Asian languages.

We recently finished work on a set of neural Sanskrit grammatical analyzer
tools together with Oliver Hellwig based on the annotations of the DCS.
This annotation system is now part of the interactive interface at
dharmamitra.org: When typing Sanskrit input into the translation box, a
button with the label 'grammar' appears below the translation box and when
clicking on this, the analyzed Sanskrit sentences become visible. This
tools currently provides word segmentation, lemmatization and
morphosyntactic tagging.

We also have inference scripts for this system on this github repository
for those of you who want to run the tools independently on their own
machine (a GPU is advisable as it might otherwise be very slow):
https://github.com/sebastian-nehrdich/sanskrit-analyzers
Among these applications you will also find dependency parsing for Vedic
Sanskrit, a function we do not yet support interactively on the website,

A publication on the architecture, data etc. used for these tools is
currently on the way.

We also are open to providing API access for individuals and projects that
would like to use these tools in their workflow. Feel free to contact us if
you are interested!

Dharmamitra.org also works on providing machine translation capabilities
for Sanskrit into English and other languages. In case you are interested
in this topic and would like to learn more, perhaps even collaborate or
contribute in some way, feel free to reach out to us. We are more than
happy to work together with people that want to explore the possibilities
of this technology.

With best wishes,

Sebastian Nehrdich
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://list.indology.info/pipermail/indology/attachments/20240829/701b72df/attachment.htm>