Thursday, November 22, 2007

IBM India develops Desktop Hindi recognition technology

IBM India Research Laboratory has developed a IBM Desktop Hindi Speech Recognition technology, which can be used to integrate the speaker independent continuous speech recognition of Hindi to edit the translated text through Shrutlekhan-Rajbhasha in the MANTRA-Rajbhasha System.

The IBM Desktop Hindi Speech Recognition technology understands and transcribes human speech with little use of keyboards. So it helps people, who are unfamiliar with computers or the Hindi language. The speech recognition technology eliminates the need to learn different keyboard mapping.

The technology has been tested and trained for variations over a large number of speakers from different regions of the country. The technology could be also useful in voice- enabling ATM kiosks and in car navigation systems. Shrutlekhan-Rajbhasha has been integrated with a number of other user-friendly features such as the facility for conversion of Text to Number , Date and Currency format. It also converts Unicode text into ISFOC fonts to enable a user to import the text to any Windows based application.

Saturday, November 17, 2007

Myth of the fastest computer

India may have launched itself on the world’s map of raw computing power with the Eka supercomputer from the Tata group, but even the world’s fastest supercomputer cannot match the processing speed of your brain.
While your computer, on an average, can execute around 100 megaflops (million of calculations per second), it can barely handle dictation.
For instance, it will take a single PC more than a few days to weeks to calculate a weather map — a task best left to a supercomputer.
Your brain, on the other hand, is able to understand multiple languages, process complex visual images, control your entire body, understand conceptual problems and create new ideas.
Scientists reveal that the brain is made up of about one trillion cells with 100 trillion connections between those cells. Estimates put the brain as capable of handling 10 quadrillion instructions per second.
Now compare that to the processing speed of the world’s fastest supercomputer from IBM at over 475 teraflops (or 475 trillion calculations per second).
What’s more interesting is that the world’s truly fastest supercomputer — RIKEN’s MDGrape-3 — will probably never be officially crowned with that title, simply because it is so specialised that it can’t run the software (the Linpack benchmark) used to officially rank computing speed.
MDGrape-3 is the first machine to break the petaflop barrier — that is 1 quadrillion calculations per second — and is three times faster than the currently-ranked fastest computer in the world, IBM’s BlueGene/L. IBM’s BlueGene/P is soon slated to achieve the petaflop distinction, though, with its machine nicknamed ‘Roadrunner’.
“Getting to the petaflop stage is a Herculean task. It involves increasing the processing speed by over 10 times, besides scaling-up in a non-linear fashion. The efficiency falls as you add central processing units (CPUs). You require to radically upgrade the architecture,” says N Seetha Rama Krishna, project manager of Computational Research Laboratories (CRL) — a wholly-owned subsidiary of Tata Sons. The Tata supercomputer has been ranked the fourth most powerful in the world.
RIKEN developed the supercomputer along with Intel, and SGI in 2006 to carry out molecular dynamics simulations. In developing drugs, pharmaceutical companies have to analyse thousands of chemical compounds to find out how they will affect the protein-bonding structures in the human body.
What takes most computers hours or days to analyse takes MDGrape-3 a few seconds. The functionality is invaluable in drug research since it can drastically cut research time involved in the development of new cures. A subsidiary of pharmaceutical giant Merck has already booked time on the machine.
Construction of supercomputers is an expensive task. To get a machine from the laboratory to the market may take several years.
In Tata’s case, however, it was done in a record six weeks. The most recent development costs of supercomputers varied between $150 and 500 million or more.
However, the Tata supercomputer cost around $30 million while the Riken one was reportedly around just $9 million. That’s partly because MDGrape-3 relies on fewer chips and less circuitry than its competitors.
Besides, Hitachi, Intel, and SGI Japan supplied the hardware and absorbed part of the cost of building the machine. One measure of the MDGrape-3’s ultra-efficient computing muscle is its cost per gigaflop (1 billion floating-point calculations per second), which Riken puts at $15. By comparison, BlueGene/L’s is $140 per gigaflop.
Also, while BlueGene/L contains a whopping 130,000 processors distributed over 65,000 nodes, Riken’s closet-sized machine needs only 4,808 chips to achieve four times its speed for certain applications. The Tata supercomputer used 15,000 processors over 2,000 nodes.
Using a supercomputer is expensive as well. As a user, you are charged according to the time you use the system what is expressed in the number of processor (CPU) seconds your programme runs, says Krishna.
In the recent past, Cray (one of the first supercomputers) time was $1,000 per hour. The use of this “Cray time” was a very common way to express computer costs in time and dollars.
Meanwhile, the next generation of supercomputing — with DNA and Quantum Computing — is already being talked about. Of course, it will take at least another decade before the new technologies will hit the work floor.

Author - Leslie D'Monte, Business Standard Online