The Indian Institute of Technology Madras (IIT-Madras) is launching the “Nilekani Centre at AI4Bharat” to advance Indian language technology. According to a press release by the Institute, the centre will focus on doing foundational work to impact society at large. The team hopes that start-ups and other industries working on Indian language technology will benefit from the datasets, tools, and pre-trained models being developed at the centre. The idea is to energise the ecosystem to do more for Indian languages.
As part of its launch, which was held earlier this week, IIT-Madras held a workshop for students, researchers, and start-ups to discuss the resources available to build Indian language technologies.
AI4Bharat is an IIT-Madras initiative to build open-source artificial intelligence (AI) for Indian languages. Over the past two years, the research team has made several contributions to Indian language technology including state-of-the-art models for machine translation and speech recognition.
Speaking at the inauguration of the centre, an industry expert noted that the government’s Digital India Bhashini Mission was been launched to make all public services and information available to citizens in their own language with collaborative AI at the core of the design. AI4Bharat will further contribute to and accelerate the India language AI work as a public good and is fully aligned with the goals of the Bhashini Mission.
A representative from IIT-Madras stated that given the rich diversity of languages in India coupled with a rapidly expanding digital world, it is important to make significant advances in language technology to benefit the common man. While language technology has significantly improved for English and a few others, Indian languages are lagging. The centre aims to bridge the gap. Another official claimed that the centre is advantageously situated at the intersection of academia, industry, and organisations working for the public good. This allows its contributions to be broad-based, spanning cutting-edge AI research, open datasets as infrastructure, and applications for use by people.
Building AI technologies for the diverse set of Indian languages is expensive given the need for large datasets and computing power. In this context, it is significant that several organisations have been supporting the efforts to build open-source AI.
In May, the Ministry of Electronics and Information Technology (MeitY) held a brainstorming session for researchers and start-ups to formulate a strategy for the Bhashini platform, launched as part of the Digital India Bhashini Mission. As OpenGov Asia reported, Bhashini will ensure citizens can access digital government services and information in their native language. The platform is interoperable and will make AI and natural language processing (NPL) resources available to MSMEs, startups, and individuals.
Essentially, the initiative will create and nurture a digital ecosystem that involves central and state government agencies and start-ups developing and deploying electronic products and services in Indian languages. Through Bhashini, the government intends to increase the amount of content in Indian languages on the Internet, especially in domains of public interest, like governance, policy, science, and technology. Initiatives arising out of the confluence of AI techniques and NLP, like speech-to-text technology, will increase the reach of governance. As public websites become multilingual and interactive, the reach of public welfare schemes will also increase.