IIIT-H Played Key Role in Building Adi Vaani

Hyderabad: The IIIT-H played a central role in developing Adi Vaani, the country’s first AI-powered translator for tribal languages, launched by the Ministry of Tribal Affairs earlier this month.

The app, which currently supports Santali, Bhili, Mundari, and Gondi, aims both to preserve endangered tongues and to make governance, education, and healthcare more accessible in remote tribal areas.

The initiative, launched under Janjatiya Gaurav Varsh to commemorate the 150th birth anniversary of Birsa Munda, was developed by a consortium led by IIT Delhi, with contributions from IIIT-Hyderabad, BITS Pilani, IIIT Naya Raipur, and several Tribal Research Institutes.

At IIIT-H, the Language Technologies Research Centre has been leading the technical development since mid-2024. Faculty members Prof. Radhika Mamidi and Prof. Anil Vuppala, along with research scholars, built two-way translation systems between English, Hindi, and Santali using Transformer-based sequence-to-sequence models.

Native speakers were integral to refining the translations. “We realised that for low-resource languages, accuracy cannot come from machines alone. Native usage had to anchor the models,” Prof. Mamidi said.

Alongside translation, IIIT-H developed text-to-speech tools for Santali, Mundari, and Bhili, with Gondi currently under development. “Working with tribal speakers who recorded thousands of samples at our labs showed us how closely technology and community have to work together,” a researcher said.

The institute’s Product Labs team also handled the deployment of Adi Vaani on the cloud, making it accessible through both Android devices and a web platform. “Accessibility was as important as accuracy. People must be able to actually use the app in villages and schools,” Prof. Vuppala explained.

Beyond translation, Adi Vaani has wider applications. It can digitise folklore and oral traditions, create subtitled awareness videos in local languages, and serve as a teaching aid for younger generations. “Our aspiration is to eventually bring NCERT books and government scheme information into these languages,” Prof. Mamidi added.

Highlighting the scale of the challenge, the researchers pointed out that India is home to 461 tribal languages and 71 distinct mother tongues, of which 42 are critically endangered. By combining advanced AI with grassroots validation, IIIT-Hyderabad’s work on Adi Vaani marks an important step in ensuring these voices are not lost.