| Latest CRM News |
| Research Reports |
| Products & Services |
| Business Deals |
| Corporate Orders |
| Corporate Performance |
| HR Watch |
| Submit your Story |
| Academic Papers |
| Articles |
| Case Studies |
| Presentations |
| White Papers |
| Research Reports |
| Finance |
| Retail |
| Telco |
| Government |
| Healthcare |
| Utilities |
| Editorial |
| Highlights |
| Experts Corner |
| Experts Panel |
| Ask the Experts |
| Books |
| Free Membership |
| Corporate Membership |
| CRM Software & Systems |
| Professional Services & Consultants |
| Analyst Groups & Research Services |
| Resources & Associations |
| Exhibitions & Conferences |
| List your Company |
| Home | | News | | Events | | Careers | | Library | | Topics | | Members | | Vendor Directory |
Nuance’s SmartListener Technology Boosts Speech Recognition Automation
Nuance today announced Nuance’s SmartListener™ software technology offering businesses a step forward in speech recognition accuracy. SmartListener technology gives businesses the confidence to automate more solutions in their call centers by meaningfully improving application automation and the overall caller experience.
By helping speech solutions proficiently handle unexpected caller inputs, SmartListener technology can increase the savings from speech applications by quickly and cost-effectively achieving greater automation rates. SmartListener technology can help call centers obtain more out of their investment in legacy and new speech solutions by addressing the single biggest hurdle to flawless speech recognition: out-of-grammar inputs.
Callers don’t always provide answers to speech solutions that the system has been designed to understand. Most people have used a speech system that offers the caller choices, such as "please say yes or no" or "you can say statements, technical support, or billing." Nuance data indicate that many times callers provide answers that sound reasonable, such as "I’d like billing please," but that are not in the precisely modeled list of acceptable responses known as the grammar. These out-of-grammar responses are therefore not accurately recognized, causing issues such as:
The system asks the caller to retry answering the same question
The system transfers the caller
When callers are not recognized or who are asked to repeat themselves frequently understandably become frustrated. By using SmartListener technology, a speech-enabled automated system is able to accurately recognize many out-of-grammar inputs, meaning that callers are recognized correctly the first time more frequently. Consequently they need to repeat themselves and transfer to agents a lot less often.
"Out-of-grammar error rates present a set of challenges to businesses and their customers," said Steve Cramoysan, research director, Gartner, Inc. "Innovation that enables speech solutions to accurately interpret the meaning of the unexpected words and phrases that customers use will help drive the ease of use, quality and accuracy of automated speech solutions while improving the most important element - the customer experience."
By analyzing Nuance’s Deployment Databank™ Nuance discovered that these out-of-grammar utterances are the primary cause of callers failing to stay in an automated customer service solution they have started using. In fact, out-of-grammar utterances outnumber other performance issues by a factor of up to 5:1. Such issues lead to customer frustration and business inefficiencies.
"Addressing out-of-grammar represents the greatest opportunity for a profound improvement in speech solution performance," said Lynda Kate Smith, "Out-of-grammar utterances are the largest single challenge facing speech applications today. SmartListener technology significantly advances how the industry can improve the caller experience while providing cost savings to business."
SmartListener technology was achieved through the combination of Nuance’s research team and their access to the Nuance Deployment Databank According to the speech science analysis of the Nuance Deployment Databank - consisting of terabytes of data and metadata gleaned from real world examples from around the globe - when the caller replies with responses for which the application has been designed, accuracy is typically well above 90%. The accuracy rate when a caller inserts responses for which the application isn’t designed, drops to roughly 16 percent. SmartListener addresses this problem by enabling speech solutions to understand what the caller intended, even if their response doesn’t precisely match the application design.

