Getting good customer service from an interactive voice response (IVR) system while in a noisy room is next to impossible for most callers.

Interference from airport chaos, honking horns or howling wind is enough to trip up even the best IVR’s and send customers away even more frustrated than they were when they first made the call.

But, thanks to Microsoft’s deep neural networks (DNN) technology and [24]7's customer engagement platform, customers now have a 25 percent better chance of being heard accurately.

New Technology

With the newest release of its customer engagement platform, [24]7 is taking advantage of one of Microsoft's 2012 acquisitions. It's using Microsoft’s enterprise IVR business to integrate DNN technology into enterprise IVR’s, raising the expected rate of accuracy to 95 percent.

“Despite advances in speech recognition for consumer apps like Siri, Google Voice Search and Cortana, speech recognition for customer service has not really improved over the last decade,” said Patrick Ngyuen, CTO for [24]7.

“DNN is the new technology that allows a large population of users to speak naturally and be recognized accurately, even under difficult conditions such as noisy environments and foreign accents.”

Misunderstandings Abound

Patrick-Nguyen
According to Ngyuen, IVR’s in the United States typically misunderstand callers between 5 percent and 20 percent of the time, depending on the question.

Outside the United States, error rates can jump as high as 30 percent, he added.

“Many of these errors are caused by noisy environments ­– including background noise or poor phone connections – or non-native accents,” he explained.

“DNN speech recognition systems are trained on much larger data sets than the prior technology, and therefore include a much more representative set of speakers and environments.”

For example, Microsoft’s DNN is trained on more than 10 billion utterances from Bing Search, Xbox and Windows Phone, he said.

With the ability to recognize customer responses more accurately, this technology can help companies deliver better customer service, and a more positive customer experience, overall, claims a release by [24]7.

In addition, customers will feel more comfortable turning to a company’s self-service channel for help, and leave more satisfied with the interaction.

Looking Ahead

Will self-service replace live representatives?

Although customer service representatives are here to stay, Ngyuen believes improving voice recognition will help free them up to address more complicated customer issues.

“Many studies have shown that customers prefer self-service, as long as the self-service works,” he concluded.

“By improving recognition accuracy through DNN technology, more customers will adopt speech self-service, and fewer will see the need to ‘zero’ out or ask for an agent. As a result, using speech to interact with automated customer service systems will become commonplace in mobile apps, at home, and in the car.

“Customer service representatives won’t go away; they’ll be less burdened by common requests and can turn their attention to more ‘interesting’ requests.”

Creative Commons Creative Commons Attribution-No Derivative Works 2.0 Generic License Title imafe by wtbyan2.