Man Vs. Machine-New Steps Taken To Boost Usage Of Speech Recognition
Most members still aren't ready to talk with a machine.
That's what Travis Credit Union has learned since it implemented speech recognition technology on its call center menu, according to Craig Crismon, assistant vice president of Information Technology at the $1.4-billion CU.
Instead, most members continue to use Travis CU's touchtone option to navigate the call center, even two years after the speech recognition system was added, Crismon said.
"We've only had 10% to 15% of our calls move over to the voice recognition," he said. Travis fields about 45,000 calls through the call center every month, he said.
Speech recognition allows members to speak to a call center telephone system as if conversing with a live agent, instead of listening to and punching through an elaborate menu of touchtone options.
Members just don't seem to get it, Crismon continued. "The problem is in educating the membership. We need to get the message across to the members that they can simply say 'I want to make a $500 deposit into my checking account.'"
Travis has attempted to alert members to the benefits of voice recognition by publishing articles in its quarterly newsletter, Crismon noted.
Ute McKinnell, assistant vice president of central operations at Travis CU, thinks low usage was complicated by the fact that certain English language accents are not recognizable. "English is not my first language and the system has trouble understanding me," McKinnell said.
Perhaps accent recognition will improve - last month, Travis CU switched to a new version of the voice recognition technology, provided by Wixom, Mich.-based Maxxar Corp. Maxxar offers computer telephony solutions to financial institutions.
Maxxar's new Total Natural Transaction (TNT) platform incorporates "major enhancements over our first generation speech recognition product," said Max Bishop, vice president of Business Development at Maxxar.
"The greatest advances are in the areas of user convenience and human factors," he continued. "We strongly believe that these advances will result in greatly improved user acceptance and satisfaction."
TNT is "very good with most accents," Bishop said. "TNT can adapt to accents and colloquialisms by collecting low-confidence level utterances from callers and then adding them to the recognition vocabulary." Graphical tools then report how well the system is adapting to accents and colloquialisms, he said.
In any case, voice recognition adoption at Travis Credit Union isn't dependent solely on technology or member education.
"Another challenge is that we left our touchtone system up when we launched the voice recognition," said Crismon. "The path that we've taken is to offer both services."
The 123,000-member CU believes that the natural voice recognition option, though less popular, is still worth offering as a service to members, Crismon explained. "Members who use speech recognition seem to like it," said Crismon. "Once a member realizes that you can simply say a phrase and get an answer without needing to navigate a menu, they love it. Our goal is to make self service as viable as possible."
TNT includes speech recognition for American English and Latin American Spanish as well as touchtone technologies and text-to-speech. Travis will soon take advantage of additional options, including speech-enabled lending, Crismon said.
The 11-branch Travis CU runs three TNT-driven boxes, enabling a total of 72 telephone lines, Crismon said. Maxxar is a subsidiary of data-processing application provider Open Solutions, Inc., Glastonbury, Conn.
For info on this story:
* Travis CU at www.traviscu.org
* Maxxar TNT at www.opensolutions.com/ifsCU-tnt.htm www.visualcalc.com