| | | | | |       Listen to Me       | | | | | |

Speech recognition has been a long time coming, but its uses are proliferating before our very voice boxes.

By the Numbers
Cognitive Speech	Voice recognition has failed to live up to the hype, but the automotive, telecom and videogaming industries have become active users.
97%	Estimated growth for the voice-recognition market by the end of 2004.
35	Years since the premiere of Stanley Kubrick's movie 2001: A Space Odyssey, starring a chatty supercomputer named Hal.
2013	The year dictation accuracy is expected to rival human accuracy.
95%	The current accuracy rate for better speech-recognition engines.
Sources: Datamonitor; Microsoft; Giga Information Group.

A house for sale in Long Beach, Calif. can accumulate ten bids the first day it is put on the market. Realtor John Johnson is using a new speech-recognition service to speed his clients to the top of the list. If he passes a For Sale sign, he immediately dials a number from his cell phone and speaks the address. The virtual woman's voice on the other end will respond with the asking price, say, or the square footage. If he supplies a client's criteria such as desired number of bedrooms and neighborhood, the service will ring him with appropriate new listings. Though the service's fee is $30 a month, Johnson expects the technology, created by a new firm called Newport Wireless, will help him close 50% more deals, earning him an additional $70,000 each year.

The old joke about speech recognition was that, for 20 years now, the boom was two or three years away. The industry was tarnished by the accounting scandal and ensuing bankruptcy in 2000 of Lernout & Hauspie, which sold the well-regarded Dragon technology to ScanSoft in 2001. In September General Magic, which created software applications for voice recognition, went out of business after failing to raise additional financing. And voice-driven information providers such as Tellme blew through tens of millions of dollars before realizing that selling voice services to companies was far more promising than selling ads spoken to consumers over the phone.

But now plenty of companies are finding smart ways to apply voice-recognition technology to cut costs and improve services. While few expect speech to replace the keyboard as the interface for basic computing, it works well in specific niches such as mobile access of important data and automating call centers for the travel industry, financial services and phone directories. Among the hundreds of companies hearing voices with a computer are AOL, FedEx, Honda, Sony, Sprint, T. Rowe Price, United Airlines and Verizon. The voice-recognition market should hit $695 million in 2002, up 10% from 2001, according to Datamonitor.

Faster chips and smarter algorithms are making the case for recognition ever stronger. Since 1985 the typical vocabulary recognized by mainstream speech software has grown a hundredfold to 100,000 words, while the average price has decreased from $5,000 to $50 in the same period, according to Ray Kurzweil, a creator of speech-recognition software and author of The Age of Spiritual Machines. Accuracy has gone from poor to good.

Julie, Amtrak's female virtual agent, has been talking with rail passengers since October 2001 about train arrivals and departures, and since last June about schedules, fares and reservations. By May of this year Amtrak expects she will also handle credit card ticket purchases. Julie, a product of SpeechWorks software and Intervoice hardware, has increased customer satisfaction by 45%, so that now 13 out of every 50 customers get their information before hanging up. Not great, but Amtrak's earlier touch-tone system, since phased out, satisfied only 9 of every 50 customers. Amtrak says the $4 million it cost to buy and develop Julie paid off in the first 12 to 18 months by cutting down on the need for more human agents. Likewise, British Airways is saving $1.5 million a year using technology from Nuance Communications to run its flight information service.

In August Sony Computer Entertainment America introduced Socom: U.S. Navy seals, its first PlayStation 2 videogame that allows players to speak commands like "Deploy grenades" to eliminate enemies. The $60 game, which uses ScanSoft speech technology, was one of Sony's bestsellers last year, with 450,000 sold. It had a better first week than any other PlayStation 2 title Sony has released itself. Sony says sales could have been double if there were more headsets on the market.

Girls who buy Radica Games' Girl Tech Password Journal ($20) don't have to worry about younger brothers prying into their private lives or losing a key. The journal unlocks when its embedded chip recognizes its owner's voiced password. If a girl is hoarse from a cold, the security can be taken down a notch. It's Radica's bestselling product in the U.S.

Luxury cars from Infiniti and Jaguar have used voice recognition for several years to control the radio, temperature and navigation systems. But now the technology is starting to trickle down into mainstream models. The 2003 Honda Accord has added voice recognition from IBM's ViaVoice as part of a $2,000 navigation system. A fifth of Accord buyers since November have opted for the voice-navigation system, compared with the 5% who have opted for a similarly priced DVD player for the rear passengers.

Dr. Nathaniel Soper, a gastrointestinal surgeon in St. Louis, has used voice recognition in 50 to 60 minimally invasive laparoscopic surgeries in two years. With technology provided by Computer Motion, Soper speaks into a headset to direct the robotic arm guiding the laparoscope into the abdomen. Soper, one of 3,000 users, says the robot isn't perfect. There is a slight lag between his command and the response, a flaw that wouldn't occur with a trained human holder. But it cuts down on the hubbub in the operating room, and people are not always available.

The VA Medical Center in Memphis has invested $277,000 in Dragon software to allow its doctors and nurses to dictate medical records into the computer network. One-third of the staff, or 450 clinicians, are now using Dragon. Doctors can access electronic records throughout the hospital whenever they want, rather than wait 24 to 48 hours for a transcription. Even though the yearly fee for the software and its support has gone up three times this year because of a contract change, the hospital still expects to save at least $60,000 a year by cutting down on use of the transcription service.

This month a new company called VoiceBox Technologies is introducing PC software that fetches voiced requests for stock information or Beatles songs from the Internet. Later in the year it will begin selling a Frisbee-size device that will transmit spoken directives wirelessly to a PC. Future versions linked to a TV might be able to record a 9 p.m. episode of Will & Grace. "There are certainly a number of companies who have found specialty niches to apply this to," says Michael Kennewick, VoiceBox's chief executive. "The market's ready."

Copyright © 2002 Forbes.com. All rights reserved.