How will the government’s new AI enhancements improve multilingual digital interactions on the Bhashini platform?

To empower users to engage digitally in their native languages, the Ministry of Electronics and Information Technology (MeitY) is enhancing the Bhashini language translation platform by developing a generative AI-based conversational system that supports multi-language input in both voice and text formats.

For this initiative, the government aims to leverage the current Bhashini platform, powered by artificial intelligence, and elevate it with advanced Natural Language Processing (NLP) techniques. This will ensure precise translations and provide users with clear, contextually appropriate responses.

Bhashini’s objective is to create a national public digital platform for languages, improving access to content and services in various Indian languages through AI and NLP technologies.

In a recent Request for Empanelment (RFE) tender issued by the Digital India Bhashini Division (DIBD), an independent division under MeitY’s Digital India Corporation, the government seeks to consult with industry experts for the empanelment of system integrators to develop the necessary platform. Feedback from potential bidders will be gathered until July 17.

IIT-Madras’ lab AI4Bharat has launched the IndicVoices dataset, covering 22 languages, contributing to this initiative.

The proposed system workflow includes features such as a denoiser to reduce background noise in voice recordings and a gender classification module to identify the speaker’s gender. It will also automatically detect the language from audio to enable appropriate translation, and feature automatic speech recognition to convert spoken language into text. Other features include neural machine translation for text translation between different languages and a profanity filter to exclude abusive words.

To increase Bhashini’s adoption nationwide, MeitY plans to collaborate with educational institutions, incubators, and accelerators, seeking technical proposals from interested parties. Selected institutions will become regional nodal centers promoting Bhashini and will conduct research to improve machine translation accuracy and fluency, as well as develop domain-specific language models for sectors like government, education, and healthcare.

Bhashini also plans to introduce gamified mobile applications and educational tools for its promotion, with support from selected collaborators in their development.