Home > Rediff Guide To The Net > Features
At your service
January 22, 2003 18:59 IST
The Samaachar Vaani will scour the Web to bring you the latest news and even read it out to you.
If you hate browsing endless pages of news on the Internet looking for what you need, this one's for you.
Rohit Kumar, a student of the Punjab Engineering College (PEC), Chandigarh (under the guidance of Kishore Prahallad of the Language Technologies Research Center, IIT, Hyderabad) has developed a newsreader software called Samachaar Vaani. It downloads the latest news from the Internet and converts headlines and stories to speech. At the moment, it is available for Hindi content.
It is an on-going effort of this Center and aims to develop language technology products that will enable people to access any information on the Net by making it available in their native language. Rohit was a summer intern at LTRC when the idea for Samaachar Vaani was conceived, and later took it up as his minor project at PEC. The software is based on the unique text-to-speech synthesis system (TTS) developed by the LTRC.
Rediff Guide to the Net catches up with Rohit Kumar to find out how this personal news messenger will work.
How does Samachaar Vaani work?
The newsreader has two components:
1. News Server: News service providers deploy the news server on their Web server and we make minor configurations to it. As a result, the news service provider is able to provide news not only as text but also as speech, to the user.
2. News Reader Client: The users download this very small (~40 KB) software (currently available both in Windows and Linux). The software can be provided by the news service provider. We can customise the client to work only for that particular news service, or for specific user names. Another idea is distributing a generic newsreader online and providing guidelines to users to configure it for various news services.
What is the technology behind this software?
We at Language Technologies Research Center of IIIT, Hyderabad, have been successful in building an unrestricted domain Indian language text-to-speech (TTS) system. We are currently working on a Hindi system that fits in 1.45 MB flat. By unrestricted domain, I mean, that it can convert to speech anything in the Hindi language. It is this TTS that the application uses as its backbone.
The availability of such a small system opens up the possibility of putting an Indian language TTS on small systems like PDAs and embedded devices including mobiles. The only other system available at comparable sizes for Indian languages is the Dhvani system developed by a group at Indian Institute of Science, Bangalore. That system is more than 2 MB in size.
How is this software different from other speech synthesis software?
I don't really know of particularly any news service that provides spoken news by automatic speech synthesis as yet. Of course, there may be services like Ananova. But they are different as they have built the automated system first followed by the news service.
The Samachaar Vaani software I have developed is intended to allow existing news service providers to extend more service to the user at no extra cost, beyond the one time cost of the software and its components (the text-to-speech system).
Currently, I have adapted the software to work with the BBC Hindi site. But its architecture is such that it can easily be adapted to any other site. We can also distribute a plug-in for the software for any new site. Only, the plug-in needs to be site specific.
Are you expecting any sponsors? Does it require more funds for further research?
I am not really expecting any sponsors. But if any news service shows interest in providing such a service on their site using my software, I would like to adapt the software for them. It would all depend on the kind of offer I receive.
I have developed the Samachaar Vaani newsreader fully without any funds. I would be interested in bringing this software to perfection if somebody is interested, commercially or otherwise. And for that, I would need funds as a researcher. Currently the downloadable distribution is unavailable, as it needs some packaging. I also need additional time.
What other applications could this service have?
It depends on the creativity and requirements of the user. Instead of the software reading out news, it can read out emails, announcements in a public place or anything else as long as it is in Hindi. In future it could be in any language supported by the system.
The text-to-speech synthesis system can also find use in several other applications:
A talking tourist aid that works on handheld devices such as the Simputer. This allows a tourist to interact with a local person in his native language. The tourist can select queries or commonly used phrases (displayed on the menu in English) and the PDA would speak it out in the local language such as Telugu, Hindi etc.
Some other possible applications could be announcement systems at railway stations, offices or industries; telephone-based query systems or voice response systems for banking, travel enquiry and shopping; rural kiosks with touch-screen menus based on images and enquiry systems in Indian languages.
A major impact of speech synthesis systems is for the blind or visually impaired computer users.
Samples of news reading by this software: (click on a link to hear the sound file. You will need a Windows Media Player)
News 1 | News 2 | News 3 | News 4 | News 5
Other TTS systems:
Vaani – TTS for Telegu
Text-to-speech synthesizer for Tamil (developed by IITM)
Ananova demo | Bell-Laboratories TTS
Multilingual TTS (from AT&T Labs)