An IIT-M faculty team wins the GE innovation award for developing software that is in sync with India’s multilingualism.
Every year, the Shastra festival of IIT Madras is something engineers look forward to. This year, at Shastra 2013, the GE India Innovation award, in addition to the first prize in the research expo, was given to the team led by Hema Murthy, faculty at the Computer Science and Engineering department of IIT-M for their work in conversion of text to speech. “Our most important contribution is that our software deals with regional languages, and that it is bilingual with real voices instead of synthetic speech which is used in other applications,” says Hema.
What is more, they had on board their development team right from the start a visually-challenged person who would not only provide a feedback but also would be involved in all stages. Taking into consideration that Indians are multilingual, the application is bilingual with the same voice speaking in both languages, which makes it user-friendly. This can be ported into Android-based platforms and even loaded into the Akash tablet that the government is planning to give students.
Open source compatibility
Moreover, the software will be completely open source compatible with Windows and Linux platforms. This software promises to be a boon for the visually-challenged from the underprivileged sections of society. The head of Technology Development in Indian Languages, Swarnalatha, appreciated their goal of making this technology available to people. She released the first round of funds amounting to Rs. 2.5 crore in August 2011 and the second round of Rs. 12.7 crore, recently.
They realised early on that the Indian accent was important to imitate if they wished their users to make good use of the application. So, they avoided the “convent-accented” speech. How do they spread awareness about the existence of this software? By tying up with an NGO Darshini, which is the women’s wing of Exnora, and AccessIndia.org, a forum for visually-challenged people, they have promoted its use among visually-challenged people. Initially, they drew a lot of flak from the user end. “But, we welcomed it. I firmly believe that you should listen to the user,” says Hema.
The group consists of 60 people from across the country working in the three CDACs, universities, IISc, IITs and S.S.N. College, and so on. They work with languages, including Bodo, Assamese, and so on.
The GE India Innovation Prize was given at the Research Expo event of Shastra 2013 organised by IIT-M. “We just wanted to showcase our work at the expo. But GE was there, and they were judging the exhibits and were impressed by the social consciousness in our work. When they announced that in addition to the first prize in the expo, we were to be given the GE India Innovation award, it came as a surprise,” says Lakshmi Priya, a team member. The prize carries a cash award of Rs. 1 lakh, in addition to the citation.
“Text-to-Speech conversion is important and there is a lot to be done in this field that would spill over one’s lifetime,” says Hema. In the long term, there could be voice conversion, dubbing, and it should be possible to characterise one’s voice box and get the conversion done in that person’s voice itself. There is much to be done in this field and a marriage between language science and computer science is very much needed. While it has happened in the West, it is yet to happen in India. “While Indian linguists are very much respected, their interest seems to be in exceptional and rare languages; in our field, the interest is in languages like Tamil…,” she adds.
4-day winter school
To promote research and study in this field, IIT-M is organising a four-day winter school on speech and audio processing, in which the first day will be devoted to explaining the concepts to beginners. Students of engineering and computer science may find an interesting opening in this winter school. Details may be had from http://www.iitm.ac.in /donlab/wissap2013/.