Voice biometrics to be commonplace in customer service: Nuance

Voice biometrics is one of the key technologies that can enable a more intelligent and human-like interaction with technology, and in the future, a concept like voice biometrics will be considered commonplace, according to Brett Beranek, director for product strategy at software firm Nuance.

"There will be an expectation that technology is able to recognise who we are and skip over the authentication process -- and PIN, password, and security questions would just be a thing of the past," Beranek told ZDNet.

"Then we can focus on more interesting and more challenging AI tasks such as being able to service the customer in a more effective manner; one that takes into account your emotional state, your current thought process, or even the current context that you're in."

Concerned with artificial intelligence (AI) by way of recognising a person's voice in a manner similar to a fingerprint, Nuance build voice biometric technologies for financial institutions, telecommunications providers, government agencies, insurance companies, and any sector customer service interaction occurs.

In a call centre environment, Beranek said the norm for an organisation will be to service an individual on a case-by-case basis. He said that how a financial institution or telco deals with a customer should take into account how they are feeling at that point in time.

"I think from our perspective, it's less interesting to have a computer produce the emotion -- that's not the objective. It's not to recreate humans in artificial form, but rather to have technology able to recognise that we are ourselves in different emotional states," he said.

"The system should be able to recognise that we are under duress, for example, and are able to behave differently -- that's the kind of cognitive intelligence that we want to build into the systems."

Beranek said that reading a person's emotional state could be as simple as servicing a customer after they experienced a life changing event, such as purchasing a house.

"If I call my insurance company after just being involved in an accident and they try to sell me new services, then that would probably make me very angry and frustrated -- it's just the wrong time to be doing that," he said.

"That's where I see technology being able to assist -- being able to determine that 'Hey, now's not the right time to be having x,y,z conversation'. It does help make that interaction between organisations and the customer better from that perspective."

With over 100 behavioural characteristics measured by the technology, Beranek said there is a pretty even split between physical characteristics and behavioural characteristics.

"If I'm under extreme duress, then my behavioural characteristics may be quite different than under normal circumstances and that will affect my voice and that will raise a red flag," he said. "The technology can detect that there is a significant change in behavioural characteristics."

Beranek said the technology started off in the contact centre as that is where voice-based conversations take place. Now, however, voice is being used as an interaction method beyond the phone such as within an app, as it has become more natural to use voice biometrics to authenticate individuals in those contexts.

"Voice biometrics is actually a technology that we developed in the 1990s," Beranek said.

"We had our first deployment in a financial institution back in 2001 and it was within the contact centre of a financial institution. It was specifically designed to detect fraudsters so the system would be listening in to calls and flagging the phone calls that came from known bad guys."

As far as Nuance is concerned, there are two different flavours to the technology.

The first is known as call active voice biometrics, where a consumer would be in a self-service application such as a mobile app, or even an automated system within a contact centre, and rather than entering an alphanumeric password or a PIN, the consumer would speak a passphrase to authenticate their identity.

"A lot of organisations use 'my voice is my password'," Beranek said. "The Australian Tax Office and the Department of Human Services use 'in Australia my voice identifies me'."

Although the phrase is secondary, he said it is just prompting the end user to say something that allows the technology to analyse the characteristics of a person's voice to determine if they are legitimately that person.

The second uses text independence voice biometrics during a live person-to-person conversation with a contact centre agent, analysing the customer's voice characteristics as they talk. Beranek said the technology flags the legitimacy of the call to the agent.

Differentiating voice biometrics from other biometric technologies such as fingerprints, iris reading, or DNA testing, Beranek said the latter are static and are entirely based on physical characteristics of a person's body.

"Voice biometrics is a little bit more complex from the perspective that yes, there's a physical angle to it -- your larynx and all the body parts through which the voice passes through affects the physical characteristics of your voice. But even that is variable," he said.

"If I lose a tooth, my voice will change ever so slightly; if I have a cold and my sinus is blocked, then my voice will change. So there's a level of variability to the voice even down to the physical characteristics."

Beranek believes voice biometrics will be complementary to future digital channels, with many organisations in the financial sector heading down the path of the smartwatch app. With little space on a watch to enter credentials, he said the technology will provide a very robust identity capability on such a platform.

"One of the sayings that we have in artificial intelligence is that once artificial intelligence is made available to consumers, it is no longer considered artificial intelligence," he said.

"In the 1980s if you said that there would be technology that would recognise your voice, it would be considered artificial intelligence. But once it's in use, it will be commonplace."