BHASHINI Shows Multilingual AI for Theses at ETD-2026, IIT Delhi

The Digital India BHASHINI Division (DIBD) carried its multilingual artificial intelligence tools to the 29th International Symposium on Electronic Theses and Dissertations (ETD-2026), which ran at the Indian Institute of Technology Delhi from 23 to 25 October 2026. The division works under Digital India Corporation in the Ministry of Electronics and Information Technology, and it used the symposium to make a case that a thesis written in one language need not remain shut inside it.

The session at IIT Delhi

Ms Baishali Chakraborty conducted the division’s session, titled ‘Workshop on AI BHASHINI: Bridging Divides and Increasing Productivity’. She set out how voice-first tools, speech-to-text conversion, machine translation, optical character recognition and whole-document translation can be put to work inside libraries, lecture halls and research repositories. A live run of Shrutlekh wrote down speech as it was being spoken and translated it in the same moment, a capability meant for teaching rooms, seminars and academic conferences.

How far BHASHINI now reaches

BHASHINI is built as a Language-First Digital Public Infrastructure, pooling language datasets, trained models, APIs and finished applications that other departments and developers can draw on. The numbers placed before the symposium were these:

Translating theses, not only web pages

The division put forward BHASHINI Pragati as proof that long and technical documents can be handled. Pragati is a translation workbench it built for the Comptroller and Auditor General of India, and it copes with files that run from 100 to 500 pages, with quality checks and data-security safeguards written into the process. The same arrangement, the division suggested, could be turned on research collections such as Shodhganga, the thesis repository that INFLIBNET maintains, and on university and college libraries, so that a dissertation keeps its structure, its technical vocabulary and its meaning when it moves into another Indian language.

Other tools on display

Optical character recognition was shown pulling text out of scans and photographs. SHABD UDYAN manages terminology, and transliteration keeps technical terms and proper names intact when the script changes. BhashaDaan, the citizen-contribution programme, gathers speech and text, transcriptions of audio and text lifted from images so that India’s language resources keep growing. Granthika was shown as a tool for digitising manuscripts, books, folk tales, literary work and oral histories. Asked from the floor whether machine translation can be trusted for specialised scientific writing, Ms Chakraborty replied that translation trained on a subject has to be paired with human checking. Harshad Sharma, Senior Manager (Finance) at the division, said carrying research into Indian languages is a step towards a knowledge system that leaves fewer readers out, and invited universities, libraries and repositories to work with BHASHINI on their own priority uses.