..| .. |
May 1-15, 1999 PC USER |
|||||
|
SPEECH RECOGNITION Command And Control Through years of intense research, quality voice recognition software is now available for a wide variety of applications for the home, office or mobile user. By Atanu Roy
Only a year ago, the software was still a little rough. You'd dictate, "Two turntables and a microphone." It might come out, "Two torn labels and an ice cream cone." But now, demos routinely have someone speaking faster than a news-radio DJ. The software captures it easily while rarely getting a word wrong.
Even in India, vendors have started making noise about speech recognition. That whoever comes first will get an early lead, is possibly not much the case, though. IBM's OS/Warp operating system, that the company launched three years ago, was speech-enabled, but found little acceptance among PC users. Of course, it was more due to Windows bulldozing than the 85 percent correctness of IBM's speech recognition capability. Now HCL Infosystems Ltd is pushing aggressively the Dragon Systems products, its new principals. Last fortnight, the HCL Group flagship company introduced in the market "the country's first speech recognition-enabled computer", a Pentium II 350 MHz run Busybee 2000 range PC. Priced at Rs 50,500, the new PC is bundled with Dragon Point & Speak speech recognition software. A. Mohan Rao, chief executive, Frontline Divsion, HCL Insys, points out that the companiy has specially customised the package for Indian users, and hence, would support the Indian-English accent and vocabulary. "Its vocabulary is adapted to regional names, locations and proper nouns, and would recognise typical Indian names such as Krishna, Lucknow, Chennai, Gram Panchayat, Coimbatore, Meenakshi and so on," Rao claims.
It's difficult to understand, however, why HCL Insys bundled the speech recognition package on Pentium II, when Pentium III is available. Hardware has been probably one of the biggest barriers in speech recognition. Intel had announced even a year ago that its new generation of processors, the Katmai, which is a follow-on to Intel's MMX Technology, would speech-ready. Among the most important new instructions in Katmai is a new memory-streaming architecture. Katmai New Instructions will enhance the performance of speech recognition applications, which continue to grow in use and popularity. Katmai New Instructions will speed up the front-end audio processing, and depending on the exact code used by the speech software developers, will increase the throughput of the search algorithms involved in pattern matching. For continuous speech recognition, the end result will be a reduced error rate and/or a shortened response time. This will be an exciting, enhanced capability as speech recognition is integrated in a growing number of business and consumer applications. The Atmosphere is Vocal From this point, some experts say, it should be only three or four years before speaking becomes an everyday way to interact with computers or computerised devices. At first, that sounds pretty cool. But for a whole generation of professionals who have built their working lives around using a keyboard, mouse and icons on a screen, things are going to get scary. You are about to become your father. He never liked the keyboard. He didn't grow up using one. When computers came along and he had to make friends with typing, he always looked awkward, probably pecking with two fingers and making tonnes of mistakes. If he really had to think, he used a pen and paper. It's your turn. Speaking is certainly more natural than tapping buttons lined up on a rectangle of plastic, so the transition might be easier. Then, while you might learn to speak to your computer, you might never learn to think that way. "The creative process is changed," says Joel Gould of Dragon. "You have to learn to think with your mouth." At some point, you will be able to make yourself look old by admitting you know how to type. Keyboards aren't likely to go away. They'll stick around as backup ways to put in information. But there will be no incentive for anyone to learn touch-typing. It will become the equivalent of knowing how to use a slide rule or tune in a TV station using rabbit ears. You might be able to hide your keyboarding skills, but at some point, in a crowded room, you will see a mistake on a document and call it a typo. The office won't be the same. Gotten used to acres of Dilbert-style cubicles? "If voice becomes the interface to the standard computer, it would definitely impact cube life," says Gregg Armstrong, vice president at Starfish Software. The office will become a yak-fest. Human voices pouring over cubicle walls would be distracting. She figures cubes will become more enclosed and podlike. "We'll all be outfitted like telemarketing people!" Speech Rules for the New World New rules will have to be imposed for meetings. The young pups in the office will bring in their hand-held computers that have voice input. They will want to use them to take notes. "You could not have everyone talking into their computers," says David Nahamoo of IBM Research. You'll get mad at your kids' teachers. Your parents probably couldn't understand how they could let you use calculators in school. They decried the erosion of basic math skills. Well, you're going to do the same over spelling. If computers almost perfectly capture speech, why should kids dwell on spelling?
You'll never get over people talking to their microwave ovens. After putting speech technology on powerful desktop computers, the next step will be to put speech into many devices that use a computer chip. The VCR, cell phones, ATMs, thermostats, copiers-eventually all could be voice-guided. In the kitchen, it won't be such a strange thing to tell your microwave oven, cook this for 3 minutes, Nahamoo says. Computers also need to be taught to be cautious. Imagine after the after bitter altercation with you, next day morning your subordinate enters the office and shouts in a full-throated, deep baritone, "All hard disks, reformat!" Most probably, computers will talk back. "Down the road will be conversational technology," says Bob Kutnick of Lernout & Hauspie. The computer would not only recognise but would understand what you're saying and either ask questions or reply. It's not science fiction. It's research and development. By 2009, predicts speech pioneer Ray Kurzweil, automated personalities will take care of routine business transactions like flight reservations. By 2019, he says, computers should be so good at conversation, a person could have a full and rewarding relationship with the latest product from Dell or Gateway.
|
|
Issue Contents Write to us Subscriptions Syndication INDIA
TODAY | BUSINESS
TODAY | INDIA TODAY PLUS | TEENS TODAY © Living Media India Ltd |
 |
 |
Once considered the realm of science
fiction writers, computers that process spoken commands and translate human dictation are
now a reality. Speaking to your PC with voice commands is no longer the fodder of movie
dreamers. After decades of trying, three companies-Dragon Systems, Lernout & Hauspie
and IBM-are selling remarkably good software that lets you speak into a microphone to
dictate documents and control your PC for a wide variety of applications for the home,
office or mobile user. 
The major speech companies-which include IBM,
Dragon and Lernout & Hauspie-are locked in a heated battle to dominate the PC space.
But since Dragon abandoned the market, none have expressed serious interest in the Mac. 
That might not be as terrible as it
sounds. In math, Nahamoo notes, calculators let teachers shift from focussing on
computation to emphasising problem-solving, a higher-order kind of math. Speech software
might let teachers dwell less on mechanics and more on creativity. 
