[Indic-computing-users] NEWS: IBM to unveil speech recognition computer tech for masses

[Frederick N. <fr...@by...>]

IBM to unveil speech recognition computer tech for Indian masses

By Sumeet Chatterjee, Indo-Asian News Service

New Delhi, Oct 9 (IANS) You don't have to be a Gen-X tech wizard to imagine
that computers of the future will listen to your spoken commands - that too
in Indian languages.

If the global tech giant IBM has its way, then one can soon open a Web site
or spreadsheet page on the computer screen without tapping on the keyboard.
All one has to do is speak out the command in Hindi or any Indian language.

"We expect to launch the speech recognition tool in the Indian market in the
next few months," said Manoj Kumar, director of IBM India Research
Laboratory, the eighth research laboratory of IBM Research.

Three of the seven IBM research laboratories are located in the U.S. and
four are situated in Zurich, Beijing and Tokyo and Haifa, Israel.

"Initially we are developing only Hindi-enabled speech recognition
technology. Developing other Indian languages would be a lot more easier
once we successfully launch the Hindi-enabled tool," Kumar told IANS.

According to experts, speech recognition computer technology can be very
useful in the Indian context as it provides an easy interface for
interacting with computers for those unfamiliar with computers and English.

Using such a convenient means of rendering information to or from the
machine would mean that the end-user need not be computer literate and still
can use the power of the IT industry.

"Since there is no standard for keyboard input in Indian languages, using
speech recognition for such languages would eliminate the need to know the
keyboard mapping for the different fonts and related problems," said Kumar.

He said that speech recognition technology could be a vital step in bringing
the IT world to the vast non-English speaking Indian masses.

"Currently we are working on increasing the accuracy of our Hindi speech
recognition system by enhancing the acoustic and language models. The
recognition system performs well over a trained context.

"Our goal is to cover more Indian languages and then to build a multilingual
speech recogniser for the Indian languages based on a multilingual phone
set."

With advances in speech-enabling technologies, proliferating electronic
devices and use of the Internet, the market for speech and language
technology is likely to grow from $1.4 billion in 2000 to nearly $3 billion
by 2005, according to IBM.

On the development aspect of Indian languages-enabled speech recognition
tool, Kumar said: "Building a recognition system that would help a machine
understand and transcribe the human speech has been a challenging field of
research.

"For any given language, the computer first needs to learn the sounds of
spellings in various contexts from a huge corpus of speech and the
corresponding transcribed text."

Set up in April 1998, the IBM India Research Laboratory, located inside the
sprawling campus of the Indian Institute of Technology here, is working on
design and prototype of the e-governance solution, Middleware.

Middleware will enable governments to deploy applications quickly at low
cost, said Kumar, adding the solution can be customised to address the
unique needs of individual state and local governments.

"Using Middleware, government processes can be updated automatically without
requiring substantial reprogramming effort and new policies can be specified
in an instructive language by a non-IT person," he said, refusing to
disclose the timeframe for the product's launch.

Kumar said that the Indian research wing of IBM was also doing significant
research and development activities in the emerging field of
bio-informatics. "It is one of the biggest areas of our research
activities."

--Indo-Asian News Service