Translate The Content in Your Local Language

Monday 4 March 2013

CDAC’s Nayana is a boon for blind people

In what could be termed a path-breaking technology development, the Language Technology division of the Centre for Development of Advanced Computing (CDAC) is in the process of giving finishing touches to ‘Nayana,’ that would help visually impaired people to read printed documents in Malayalam.

This should become a God-send for the blind when the software for text-to-speech (TTS) conversion of Malayalam is improved upon by integrating it with a hand-held scanner. If everything goes according to plan, a visually-challenged person holding a scanner above a printed page should get a voice-output very similar to someone reading out a passage for them.

Similar. Yes. But the machine-like robotic sound of the narration is yet to be corrected. “We are trying to bring in the intonations and the modulations so as to give the voice a more human-like quality,” said senior scientist and Language Technology head V K Bhadran.

Nayana can differentiate between images and text on a scanned document and recognise just the text by a Malayalam Optical Character Recognition system. The system, also developed by CDAC, recognises characters in the old script as well the new script in Malayalam.

Once the text recognition is completed on the scanned page, the Text-To-Speech system, which transforms the text into synthetic speech, is chosen. The TTS depends on a database of recorded sound segments and syllables for a voice-output. The CDAC team has tweaked the Malayalam TTS in such a way so as to incorporate some frequently used English words as well.

Other than scanned documents, a small hand-held scanner that can store a few scanned images would help blind persons to get a voice-output of even books. The team is now trying to make the scanner more user-friendly for a blind person.

“We have developed screen-readers with sound-backups that would let a blind person know if he is typing right. If we could bring in such sound-backups for the scanner, at least to identify the beginning and the end of the text, then this would work out fine,” said senior engineer Sajini.

The clarity and speed of the synthetic speech is dependent on the engine that is used - whether the TTS uses Esnola, Flite or Festival. The scientists of the Language Technology Division said that the blind people were not so much bothered about the machine-like narration as they were about speed.

“They learn quickly, grasp things easily and need texts to be read out quickly. Speed and clarity often don’t go together. The Festival-based TTS has a good memory, has been integrated with Open Source readers and the rendering is clear but many prefer Flite-based TTS as it is fast,” said Sajini.  The Language Technology Group at CDAC, which had established a Malayalam Resource Centre, is known for having developed a Malayalam Braille Editor; Ezhuthachan - a Malayalam tutor; Aacharyan - an English tutor in Malayalam and a Malayalam screen-reader, among others. Most of these can be downloaded free from their website.


Source : The New Indian Express , Thiruvanthapuram , 3rd march 2013 

No comments:

Post a Comment