The Voice of the Machine
UMaine gives its answering machine a little personality.

There are few people who spend much time calling corporations and government offices who don't eventually wish there was a way to chop through the forest of computerized telephone answering trees most large organizations (and more than a few small) feel is vital to their efficiency. "For accounting, press seven; for personnel, press eight; for screams of frustration, press nine. . . . " Which makes the University of Maine's computerized answering service so unusual.

"Good afternoon. Welcome to the University of Maine. To whom would
you like to be connected?" asks a pleasant female voice. Speak the
name of the person or department wanted, and the voice-recognition software immediately puts the call through. The software even learns to recognize a repeat caller's voice and accent and eventually drops the
follow-up prompt that makes sure the request was understood correctly.

"You can stump it, but you really have to work at it," says Les Shaw, the university's director of information technology. The system, made by
SDC Solutions of Manchester, New Hampshire, was installed more than three years ago and uses advanced voice-recognition software to direct callers to any department, employee,
or student who has a phone on the
Orono campus, including the phones in each dormitory room.

"I have a member of my staff from Russia, Alexie Strukov, and I called in thinking I could fool the answering system with his name," Shaw recalls. "I couldn't do it, no matter how I pronounced it."

The system arrived already equipped with an enormous file of first and last names, along with the ability to learn new ones. "We have a lot of foreign professors and staff and students on campus," Shaw says. Both he and his secretary -- "my speechmaster," Shaw calls her -- can access the system to add or subtract names from the files, and once or twice a year a list of names goes to SDC to be professionally recorded and added to the system.

If a caller asks for someone the system doesn't recognize, the call is transferred to a switchboard operator, Shaw says, but it handles about four thousand calls a day with 95 percent accuracy. He says some callers still prefer to talk to a human being, or they mistake the answering system's voice for a real person and try to talk to it.

Shaw says an artificial intelligence that can actually carry on a conversation is still well into the future. But at least the current system listens.

You can find this article in the September 2006 issue of “Down East: The Magazine of Maine” on page 13. To read more, go to http://www.downeast.com/index.php/page/issues/id/2474.

For more information about SDC Solutions, visit www.sdcsolutions.com.

For more information about the University of Maine, visit www.umaine.edu.