OCT 04, 2023
IIT Bombay software to enable live translations in regional languages inside classrooms
OCT 16, 2022
Indian Institute of Technology Bombay (IIT-B) has developed a software tool called ‘Bahubhaashak’ (polyglot) that will allow real-time translations of lectures inside classrooms in multiple languages with the help of artificial intelligence.
The Speech-to-Speech Machine Translation (SSMT) system allows translations from English to Hindi, English to Marathi and Hindi to Marathi. Developed by the Center for Indian Language Technology (CFILT) at the Department of Computer Science and Engineering at IIT Bombay, Bahubhaashak was initially funded as a pilot project by the Ministry of Electronics and Information Technology (MEITY) in 2020. It is now a full-fledged programme that is part of the Ministry’s National Language Translation Mission (NLTM) and will be deployed for monetisation.
The process of translation involves four steps. First, speech in a language is converted into text through Automatic Speech Recognition, and corrected for vocal disfluency such as fillers like "uhh" and "umm". Next, through machine translation the text in one language is converted into another language of choice. Finally, this text is converted into speech in the same language. The whole process takes between one or two seconds.
To enable translations, the programme developers scraped 60 lakh sentence pairs from various websites, including Wikipedia and government ones. Then through "supervised learning", where a correct sentence pair in two different languages is entered, machines are taught to predict the output correctly and translate.
"We are planning to provide this technology to private engineering colleges to enable them to translate English lectures into local languages. We are also talking to Bombay High Court to help them translate into Hindi court cases that are transferred to them from lower courts in local languages. We also expect a lot of interest from the tourism industry and travel portals like MakeMyTrip," Professor Pushpak Bhattacharya of IIT-Bombay told The Hindu.
The tool is available as a web service and there is also a plan to develop a mobile application that will be freely available to all, at least in the beginning, the Professor said.
The CFILT is now working on two more projects under NLTM called ‘Ishaan’, which focuses on languages from the northeast where text-to-text bi-directional translation will be provided between English and Assamese, Bodo, Manipuri and Nepali; between Hindi and Manipuri; and Assamese and Bodo. Under project Vidyapati bi-directional text-to-text translation will be available between Hindi and regional languages such as Bengali, Konkani, Maithili and Marathi.
The project was part of an exhibition during a two-day fair from October 14-15 at IIT-Delhi where all 23 IITs in the country displayed their research and development work. The emphasis of the fair was to take innovations developed by IITs to the industry so that they can be commercialised as well as to find solutions to local problems, according to Pawan Goenka, Chairperson of Board of Governors, IIT Madras.
The government support for Bahubhaashak comes in the backdrop of the National Education Policy 2020 emphasising on promoting multi-lingualism and Indian languages and recommending mother tongue or local language as medium of instruction in schools, colleges and universities. The All India Council of Technical Education has already translated 22 books of first-year engineering programmes in 12 scheduled Indian languages and Home Minister Amit Shah is scheduled to launch Hindi version of MBBS books in Madhya Pradesh on Sunday.