The Digital India BHASHINI Division (DIBD) has made a pitch to India’s universities, libraries and research repositories: use multilingual AI to open up doctoral theses, dissertations and other scholarly work to readers across Indian languages.
The proposal was put forward at the 29th International Symposium on Electronic Theses and Dissertations (ETD-2026), held at the Indian Institute of Technology (IIT) Delhi from 23 to 25 October 2026 under the theme “ETDs in the Age of AI”. DIBD, which works under Digital India Corporation and the Ministry of Electronics and Information Technology (MeitY), ran a workshop titled “AI BHASHINI: Bridging Divides and Increasing Productivity”, led by Baishali Chakraborty.
From lecture halls to research repositories
The session showed how BHASHINI’s language-first digital public infrastructure could serve higher education. Demonstrations covered voice-first multilingual AI, speech-to-text, translation, optical character recognition (OCR) and full-document translation. A live demo of Shrutlekh transcribed speech in real time while translating it, a capability the team positioned for lectures, research seminars and academic conferences.
A central idea was extending AI translation to electronic theses and dissertations, which are often long and packed with specialised terminology. The team cited BHASHINI Pragati, a document translation workbench built for the Comptroller and Auditor General of India (CAG) that handles documents of 100 to 500 pages with quality checks and data-security controls, as a model that could be adapted for repositories such as Shodhganga, the national theses repository run by INFLIBNET, and for individual universities.
The platform in numbers
- More than 10 billion cumulative AI inferences recorded
- 36 Indian languages supported for text and 23 for voice, plus 35 international languages
- A website translation plugin live on more than 800 government websites
The workshop also touched on OCR for scanned documents, terminology management through SHABD UDYAN, and transliteration to keep technical terms and names intact across scripts. BHASHINI also presented BhashaDaan, its citizen-contribution drive for building language datasets, and Granthika, a tool for digitising books, manuscripts, folk tales and oral histories.
Human review stays in the loop
Participants questioned how reliable machine translation can be for dense scientific and academic content, and whether PDFs could be translated on the fly. Chakraborty stressed that domain-specific machine translation has to be paired with human review and quality assurance before scholarly content can be trusted in another language.
Harshad Sharma, Senior Manager (Finance) at DIBD, said making research accessible across Indian languages is an important step towards a more inclusive knowledge ecosystem, adding that engagement with academic institutions will be key to understanding their needs and building reliable solutions for multilingual access to research content.
BHASHINI has invited higher education institutions, research repositories, libraries and academic bodies to work with it on priority use cases, domain-specific language resources and translation workflows suited to their collections.
Why it matters
Most Indian doctoral research is written and archived in English, which limits who can read and build on it. NEP 2020 calls for higher education and research in Indian languages, and multilingual access to existing theses is a practical step in that direction. If universities and INFLIBNET take up the offer, Shodhganga’s large archive could become far more discoverable for students and researchers who work primarily in their own languages. The real test will be translation quality in specialised subjects, which is why the emphasis on human validation will matter as much as the AI itself.
0 Comments
Sign in or create an account to join the discussion.