Verizon to upgrade directory assistance and picks new voice. By Peter J. Howe, Globe Staff | September 15, 2004
So long, Eryca. Here comes Darby.
New Englanders who call Verizon Communications Inc.’s 411 directory assistance will start to hear a different voice over the next week, as the phone giant upgrades to a system that will handle more calls with automatic speech-recognition software instead using a human operator.
Since the late 1990s, the voice behind “What cities and states?” and “What listings?” has been that of Eryca Dawson, 35, an upstate New York native and graduate of Howard University who has gone on to a career in radio and professional narration.
With its switch to a new system made by Volt Information Sciences Inc. and TellMe Networks Inc., Verizon is also picking up a new female voice to complement actor James Earl Jones mellifluously intoning “Verizon Nationwide 411.” She is Darby Bailey, 33, a Salt Lake City native and actress now living in Northern California who is also the voice behind AT&T Wireless Services Inc.’s #121 information service, Delta Air Lines’ Song reservation and information system, and other automated calling services.
Today, Verizon uses prerecorded voice prompts to ask for numbers and a speech synthesizer to play back the listings, but a person hears the request and selects the number from a computer screen. Often — for callers who enunciate well and don’t have a complex request — the human operator never has to speak to the caller, creating the illusion that the whole interchange has been automated and computerized.
Over the next two weeks, as Verizon begins switching local networks over to the new system in Greater Boston and New England, more and more 411 calls will actually become automated. Initially, Verizon will use speech recognition for business and government listings. By winter it hopes to use computerized speech interpreters to search residential listings too.
For people calling the new 411 on a Verizon phone, a big change will be hearing Bailey’s prerecorded voice repeat what the computer thinks it heard the caller ask for — “Ken’s Steak House, is that right?” If the caller confirms that the computer heard accurately, the corresponding phone number will automatically be played; otherwise, callers will hear Bailey say, “Please hold for assistance with your listings” before a live operator comes on.
Verizon continues to use Eryca Dawson’s voice in many areas for people calling in a repair order. It hopes the new Bailey-based system will get people listings a second or two more quickly and keep its 411 service a competitive alternative to looking up phone numbers on the Internet.
“Today, time is the most valuable commodity anybody has,” said Joanne Henke, executive director of Verizon’s LiveSource unit, which oversees 411. “This system enables people to have a clean, crisp experience from start to finish, and the best part of all this is that we still have talented operators searching the more difficult calls for customers.”
Using more speech recognition will also enable Verizon to cut back on paying human operators, but Henke said because of union contracts the company would reduce 411 staff only by attrition and other non-layoff measures.
Bailey spent weeks recording the standard voice prompts as well as thousands of individual business and place names for the Verizon project, including as many as 100 takes of lines like “city and state, please?” She got started moonlighting in voice prompts for TellMe several years ago when her husband, Scott, joined the Silicon Valley start-up as one of its first employees, and now does it full-time while also trying to launch a career as an independent filmmaker. Besides its work for corporate speech systems, the company also operates a “voice portal” at 1-800-555-8355 where people can ask for news, weather, and stock reports.
“It’s a test of my endurance,” Bailey said of the work behind creating the prompts, which she typically records in two hourlong sessions a day to keep her voice fresh. “I just try to stay in a helpful mode and try to stay focused and stay cheerful. When I’m recording a script for when something goes wrong, I try to make sure I sound empathetic without being saccharine, which can be really irritating.”
While Bailey’s voice will help introduce Verizon 411 callers to an increasingly automated directory assistance, the volume and variety of directory-assistance calls will still require a lot of human involvement, experts say.
John Makhoul, the chief scientist at BBN Technologies Inc., a Cambridge technology and research firm that was owned by Verizon for three years before being spun off last year, said from a technology standpoint, one of the crucial things the Volt-TellMe system does is have Bailey’s voice confirm with callers it actually understood what they are asking for.
“Confirmation buys you a lot. You can be brave and make a mistake; if the system hears ‘no,’ it goes to an operator,” Makhoul said. “Our goal is to have the technology eventually do as well as a human, but we’re not there yet. I don’t think there’s any system that can do as well as a human.”