Select your localized edition:

Close ×

More Ways to Connect

Discover one of our 28 local entrepreneurial communities »

Be the first to know as we launch in new countries and markets around the globe.

Interested in bringing MIT Technology Review to your local market?

MIT Technology ReviewMIT Technology Review - logo


Ram Prakash Hanumanthappa, 31

Phonetic transliteration software.

Developer of Quillpad

50 sounds of Indian languages had to be represented through 26 letters of the English alphabet. Ambiguities arose as letters like "d" could represent different sounds in words like "Hindi" (द) and "bada" (ड). To disambiguate this, earlier solutions were expecting the users to master a table of rigid rules of typing.

Hanumanthappa developed Quillpad, phonetic transliteration software, adopting a machine learning tool (CART) which learns to predict the sounds based on context so that the ambiguities are handled automatically and the users can type different intuitive phonetic spellings and get the desired output in their languages accurately.

It works by learning the patterns of a language rather than memorizing the spellings. This helps as there can be no one correct spelling when writing Indian language words using English letters. Hanumanthappa's technology can take the English words used in Indian language input and accurately transliterate it into Indian way of pronouncing the word. "It is a generic tool that generates training data automatically. Because of this, transliteration from any alphabet based language to any other alphabet based language (not Indian languages alone) can be achieved in a matter of a few hours without the intervention of any human linguistic experts," explains Hanumanthappa.

The next step for him was launching of Quillpad Mobile. "We had developed Quillpad basically so that users could send emails in Indian languages. We wanted to take it a step forward and use it to work with other input interfaces such as mobile keypads. We succeeded and did not have to modify the engine even by a byte," says Hanumanthappa. He developed a compression technique to fit the entire set of patterns into a small memory that would be suitable for the mobile phones.

Quillpad is now a powerful tool for predictive text input as it allows to type phonetic Roman characters into mobile devices. The transliteration engine can respond intelligently in other languages such as Tamil, Urdu, Arabic, and Hindi. The Quillpad engine can take input from sources as diverse as 102-key keyboards to the standard keypads with 12 keys on most mobile phones.

The integration of the Quillpad engine with technologies like the Nokia's Front End Processor (FEP) makes the multilanguage short message service (SMS) process seamless for S60 third edition device users. Quillpad interprets single keystrokes in as intelligent a fashion as possible, making it unnecessary to resolve the ambiguity of matching a numeric key to a letter using multiple keystrokes. When an ambiguity does exist, the Quillpad software pops up the options to choose by pressing the '*' key. Quillpad adapts according to a pattern-based prediction mechanism rather than a dictionary-based one and allows messages to be typed in Hindi, Kannada, Malayalam, Marathi, Tamil, Telugu, Bengali, Gujarati, Punjabi, and Nepali with a simple user interface.

2010 TR35 Winners

Aaron Dollar (video)

Creating flexible robotic hands

Rikin Gandhi

Educating farmers through locally produced video

Kim Hazelwood

Reëngineering software on the fly

Indrani Medhi

Building interfaces for the illiterate

Andrey Rybalchenko

Stopping software from getting stuck in loops

T. Scott Saponas

Detecting complex gestures with an armband interface

Jian Sun

Better image searches

Richard Tibbetts

Reacting to large amounts of data in real time


More Innovators Under 35: