Centre’s initiative takes tribal languages online

The Union government has developed an artificial intelligence (AI)-powered language platform seeking to include tribal communities in the country’s digital transformation by getting hundreds of tribal languages online and providing benefits to people in their own languages.

India has more than 700 tribal communities and over 100 tribal languages and dialects. Yet only a small proportion of these languages have a script, dictionary or substantial digital presence.

AdiVaani, developed under the ministry of tribal affairs in collaboration with IIT Delhi, enables translation and transcription between Hindi, English and tribal languages through text and speech. The platform currently supports Santali, Mundari, Bhili and Gondi, with Kui and Garo among the languages being taken up next. Its next phase envisages adding 10 more languages and significantly expanding verified speech recordings, with collections expected to run into hundreds of thousands of samples for individual languages.

The initiative is aimed not merely at translation but at making government services, healthcare information, educational material and other digital resources accessible to people in languages they use in everyday life, through text-to-text translation, automatic speech recognition, text-to-speech and speech-to-speech functions.

“Language cannot become a barrier to accessing the benefits of development,” said Ranjana Chopra, secretary, ministry of tribal affairs, stressing the need to ensure that technology reaches communities in forms they can understand and use.

The project also includes optical character recognition to digitise old textbooks, primers, dictionaries and manuscripts that exist only in physical form.

“Technology should strengthen the connection between communities and their own language and culture, rather than replace it,” Chopra said, describing the broader objective of using digital tools for tribal empowerment.

According to people familiar with the matter, native speakers, translators, linguists and community elders are involved in building and checking the language datasets. The project’s maker-checker-superchecker process is intended to identify subtle errors that automated systems can introduce, including incorrect suffixes, unnatural expressions and the loss of oral registers.

State Tribal Research Institutes are also central to the exercise. Institutes in Jharkhand, Odisha, Chhattisgarh, Madhya Pradesh and other states help identify native speakers, document dialect variations and verify translated and transcribed material before it is incorporated into the training corpus.

The larger objective, Chopra says, is not simply to make languages compatible with AI, but to ensure that they have a durable place in India’s digital future – allowing communities to access the state in their own languages while giving the tribals greater access to their linguistic and cultural knowledge. “Inclusion must begin with the language in which people think, speak and understand their world,” she added.