He's calm and knowledgeable. Patient, but efficient. And he makes a special effort to help the elderly and non-native English speakers.
It would be easy to like this guy answering the phone at the Port Authority. But his robot-like nasal tone dashes all hopes right off the bat.
"I am a voice-activated agent that can give you bus information," he says. "Where do you want to go?"
He's a computer system and he's only available for two weeks, as the Port Authority and Carnegie Mellon partner to test a new voice-activated system that allows callers to find route and time information for 10 bus routes, all based in the East End, during times when the Port Authority customer service line is not staffed.
The system is designed specifically to understand elderly and non-native speakers, two groups that often have trouble using voice-activated software. Until March 20, any caller can speak to the dialogue system to get schedules and bus information. The agent's fate after this time will depend on his success interpreting callers' requests.
With an estimated 25,000 callers a week, the Port Authority is happy to have the system, if only for a short while, to provide information to riders about buses after the customer service representatives have gone home. It fits especially well with the Port Authority's ridership demographics.
"We have a lot of seniors that ride our system," said Port Authority spokesman Bob Grove. "Some of them don't have the ability to get online or to a schedule rack for information."
Many voice-activated systems are not user-friendly for people who have trouble either hearing or speaking English because they've been designed using accessible subjects like university students, said Maxine Eskenazi, an associate teaching professor at Carnegie Mellon's Language Technology Institute and one of the lead researchers on this project.
"If you make a system for students, you'll never be able to use it for the general public," she said. "If we respond to what we feel are some of the more extreme speaker populations, then everybody benefits."
Eskenazi and research scientist Alan Black received a $650,000 grant from the National Science Foundation to research exactly how to design a system for these extreme speaker populations, specifically the elderly and non-native speakers. They worked with associate research professor Lorraine Levin, and graduate students Antoine Raux and Brian Langer, the voice of the agent.
In researching speech patterns of the elderly, the team started at a logical place: alumni back for a 50th class reunion at Carnegie Mellon. To test the system's speech clarity, alumni were asked to listen to a recording and write down what they heard. With those tests and recorded calls from the Port Authority's customer service line, the team spent almost three years building and honing the system.
Some of its special features include slowing down the pace of the voice-activated agent's speech if callers say "what?" or ask to hear it again. It also is designed to distinguish what people are saying when they speak in a noisy area.
The system is also instructive ---- it corrects callers' incorrect phrases or expressions in a way similar to how people learning a foreign language might learn which idioms and phrases to use. For example, if a caller were to ask for "bus times," the software would train the caller to ask for "bus schedule" next time.
Both the Port Authority and Carnegie Mellon are interested in using the system to help old people who will be in town for the Senior Olympics this summer, but they will need to find additional funding to move forward.