Speech Recognition Breeds New Industry

Article excerpt

KANSAS CITY, Mo. _ A doctor dictates his report into a computer before he is done examining his patient.

Across town, a factory worker reads aloud a string of numbers on a product; a radio strapped to his waist broadcasts the data to a far-off computer.

Both applications were devised by CompuSpeak Laboratories Inc. of Olathe, Kan., a player in the new industry being built around speech-recognition technology.

Bypassing computer keyboards and speaking data directly into computers is a new frontier along the information highway. Already, it is a $300 million a year business and it should more than double to $750 million in three years, according to the Yankee Group, a Boston-based consulting firm.

One of the major users of the technology is Sprint Corp. Since January the Westwood, Kan., company has marketed a voice card that allows users to speak their identification numbers into a computer reached by dialing an "800" number. Callers then verbally instruct the computer to call one of 10 phone numbers.

Futurists have long dreamed of other speech-recognition applications, such as using computer robots to automatically translate English to Japanese. Apple Computer several years ago developed a video that simulated a professor instructing a computer to gather data.

Such novel technology requires powerful computers and software that converts the human voice into a stream of 1's and 0's, the language of computers.

The hard part is programming computers to recognize context, when "two" is not "too," or "know" is not "no."

A speaker's voice, represented by analog sound waves, is converted into a digital format of 1's and 0's, and the data are then compressed. Once a digital pattern is recognized by the computer system, the computer "translates" that pattern into text on a computer screen.

CompuSpeak, which was founded eight years ago, and a handful of other pioneering companies nationwide are building businesses around specialized speech-recognition systems and software.

CompuSpeak's revenues have exceeded $10 million a year, and 100 employees work at the company.

Brian Adamik, director of consumer communications at the Yankee Group, said software companies were looking at speech-recognition technology to make it easier to use their products. Working with Lotus accounting software, for example, a computer user could say "open spreadsheet" and be ready to go.

Telephone and cellular phone companies also are deploying the voice-recognition technology. Besides Sprint, local telephone companies such as Ameritech of Chicago use similar technology.

Cellular phone companies see particular merit in letting motorists verbally interact with their phones instead of having to use their hands to make calls.

Advanced research facilities, like AT T's Bell Laboratories in New Jersey, for years have been researching the very tricky problem of developing computers and software that recognize continuous speech. …