• March 12, 2003
  • By David Myron, Editorial Director, CRM and Speech Technology magazines and SmartCustomerService.com

Speech Recognition Gets Smarter

Callers can now have conversations with a machine, as Nuance yesterday announced the availability of its Nuance Vocalizer 3.0, a text-to-speech (TTS) application. Organizations are using speech recognition technology to bridge the gap between speech and text. Some speech recognition systems "understand" a caller's phraseology and intelligently determine the most appropriate answer to the customer's query. Unlike recorded prompts that are used to communicate static information, the Nuance Vocalizer 3.0 TTS system can convert dynamic data stored as text in a database into speech. Although speech recognition software, specifically in dictation applications, has struggled with accuracy issues, industry pundits say accuracy levels for telephony has improved. "The dictation experience is so bad, but speech recognition in telephony is highly accurate. You normally see at least an 80 percent accuracy level for a business," says Lynda Kate Smith, vice president and chief marketing officer at Nuance. By converting written text to spoken words, Nuance's TTS software enables companies to automate the delivery of frequently changing information over the phone. The Nuance Vocalizer 3.0 is designed for key automated speech solutions, including directory assistance, information access and retrieval, account management, voice activated dialing, and email access. An example of the type of dynamic information the TTS solution can communicate is flight information, such as arrival and departure times and delays and cancellations. Nuance Vocalizer 3.0 emulates the conversational style of a live agent, even pausing at appropriate times for purposes of emphasis or clarity. The product is designed to communicate with callers from various regions and can teach itself to understand various dialects and accents. "We have an auto tuning capability called listen and learn where the system will automatically listen to the accent and tune itself so it understands the accent," Smith says. The solution can also be used as a sales tool used for targeted selling. Vocalizer 3.0 can identify callers by gender, based on their voiceprint, and sell to them. It can also push automated outbound calls. Callers can be notified of a specific promotion and if interested would be prompted to say, "Yes. I'm interested," at which point they would be routed to an agent. In the future Nuance will address the cumbersome call steering designs common with IVR systems. Most people zero out of IVR menus, Smith says, because they get frustrated or they can't remember which was the right option. "By the time you get to option number seven, you have to listen to all the options again," Smith says. "You saw this a lot in the credit card industry. The series of menus leads to a high number of zero-out rates. Typically the average zero-out rates in the IVR industry are 60 to 70 percent in the United States." By the middle of next month Nuance expects to ship its own natural language call steering application. "We say forget the menus. Let's do natural language understanding to direct people through our menu. Our system listens for keywords and gives callers the ability to say anything they want. The solution understands and figures out the best way to route callers," Smith says.
CRM Covers
for qualified subscribers
Subscribe Now Current Issue Past Issues