Bhasha Yantram(Language Machine)- AI Powered Gondi & Kolami speech and Text Translation Plat for Tribal Governance – Integrated Tribal Development Agency (ITDA), Telangana

Bhasha Yantram(Language Machine)- AI Powered Gondi & Kolami speech and Text Translation Plat for Tribal Governance

Language Barrier in Governance: Government forms, grievance portals, health sessions, and scheme communications in Adilabad are almost entirely in Telugu, which most local tribal citizens barely understand.
Impact on Health and Entitlements: Health and nutrition advice is lost in translation, and citizens miss out on welfare entitlements because they do not understand what they qualify for.
Grievance Redressal Bottlenecks: Grievances frequently go unfiled or misrecorded due to the lack of direct communication channels in tribal languages.
Dependence on Middlemen: Administration and citizens rely heavily on informal, inconsistent, and sometimes self-interested middlemen to translate and interpret information.
Lack of Scalability: Previous efforts, such as an official learning Gondi, were commendable but unscalable across more than 700 villages and hundreds of thousands of citizens.

AI-Powered Translation Platform: ITDA Utnoor partnered with IIT Hyderabad to build Adivasi Vaani, an AI platform supporting bidirectional speech and text translation for Gondi, Kolami, and Telugu.
Community-Led Parallel Corpus Building: A 30-member team of native-language teachers and coordinators compiled a structured corpus of over 20,000 sentences covering administrative, health, and grievance-related vocabulary.
Native Speech Recording: Recorded over 200 hours of authentic native speech audio across dialectal variations in different tribal habitations.
Voice-First Design: Designed the app with speech-to-speech, speech-to-text, and text-to-speech features to serve non-literate tribal citizens.
Institutional Integration & Training: Embedded the tool into ITDA touchpoints and provided structured training for health workers, teachers, revenue staff, and field functionaries.

First Structured Language Dataset: Created Telangana’s first digitized language dataset for Gondi and Kolami, forming a long-term public asset.
Interpreter-Free Communication: Enables real-time, direct interaction between tribal citizens and administrative officials without relying on third-party interpreters.
Improved Welfare and Health Access: Faster, clearer delivery of health and nutrition guidance, along with more accurate identification and enrollment for welfare schemes.
Accurate Grievance Redressal: Reduces unfiled or misrecorded grievances by allowing citizens to lodge complaints directly in their own language.
Replicable National Playbook: Established a validated, low-cost methodology of community-sourced corpus creation and LLM fine-tuning that other tribal-majority districts can adopt.