Computers and Appliances Are Taking On a Conversational Tone
Some people speak to their computers out of frustration. Soon, we’ll be speaking to them with purpose.
Voice technology has gone beyond recognition and is venturing into conversation, and several companies are introducing products at the Comdex computer trade show this year that allow computers and other information appliances to understand speech, allowing for more efficient and less mechanical interactions.
For the record:
12:00 a.m. Nov. 20, 1998 For the Record
Los Angeles Times Friday November 20, 1998 Home Edition Business Part C Page 3 Financial Desk 1 inches; 35 words Type of Material: Correction
For the Record
Computer company--A story in Wednesday’s paper (“Computers and Appliances Are Taking On a Conversational Tone”) incorrectly identified Ora Williamson and her company. She is executive director of Conversational Computing Corp. in Redmond, Wash.
Speech-recognition technology has been maturing for several years and been applied to tasks such as dictation. But combining that power with emerging “natural language processing,” a technology that teaches computers the definitions of and interrelations among words, means a person can give a computer a complicated string of instructions instead of discrete ones.
For example, Fidelity Investments last week announced it would implement a program that would allow customers to call in and talk to a computer as they would a broker, saying, perhaps, “Buy me 500 shares of IBM at the market.” Currently, automated call-in centers have more directed dialogues with people, asking the customer what they want to do, what stock they would like and at what price.
“Natural language, whether it’s used to type in sentences or for speech recognition, is destined to play a very central role in the operating system and the applications in the future,” Microsoft Corp. Chairman Bill Gates said at Comdex’s opening keynote address Sunday.
The market for equipment, software and services using speech technology in telephony alone could hit $38 billion in five years, according to TMA Associates, a Tarzana firm that follows the speech technology market.
Call centers that deal with intricate but non-subjective questions--including airline reservation systems, mutual funds and banking--provide ample opportunities for conversation applications.
Products are also being rolled out that apply to numerous personal computer programs, such as word processors, spreadsheets, calendars and Web browsers.
Eventually, voice may become as integral to computers as the mouse and keyboard.
“If I take the mouse away from a Windows or Macintosh computer, they’re rendered pretty hard to deal with,” said Robert Kutnick, chief technology officer of Belgium-based Lernout & Hauspie, which has voice-enabled all of the programs in Microsoft’s Office suite. “When the operating system is truly voice-enabled, when you take the microphone away there are going to be tasks that are fairly difficult to do without it.”
The goal is to replicate the “ ‘Star Trek’ paradigm,” in which people speak into the air and a computer responds, said Ora Williamson, president of Computing Corp. of Redmond, Wash. Currently, conversation technology is constrained by the quality of microphones and other sound-detection devices.
“Having to wear a headset to speak to a computer is like having to hold someone’s hand in order to have a face-to-face conversation,” Williamson said.
More difficult will be getting the computer to understand the rambling way people normally speak so that it doesn’t necessarily take in everything a person says, said Steve Erlich, vice president of marketing at Menlo Park, Calif.-based Nuance Communications. Nuance provided the technology behind the Fidelity system, which will debut in a pilot program next month.
“When a person is making an airline reservation, they might say, ‘I need to fly to Dallas on Thursday; um, wait, no, make that next Thursday, from Los Angeles, and a morning nonstop flight would be nice, but the traffic to the airport is really bad in the mornings, so maybe the afternoon would be OK,’ ” Erlich said. “The computer interface has to be simple enough for the user and organized enough for the computer.”
But “Jetsons”-like abilities will have to wait until at least the middle of the next decade, said W.S. “Ozzie” Osborne, general manager of speech systems at IBM Corp., which has about half the voice-recognition market in the U.S.
Programmers also have designs for conversation-enabled household appliances such as videocassette recorders and ovens. Said Osborne: “If you really get into the future, when I tell my VCR to record the next episode of ‘ER,’ it would say, ‘Oh, this is a rerun, are you sure you want to do that?’ ”