Computerworld

One in three prefer speech recognition to a real person

Advancements in open dialogue capabilities fuelling growth.

A study by research firm Callcentres.net has found that both Australian customer satisfaction with speech recognition systems, and the technical capabilities of the systems, have increased significantly over the past several years.

It found 58 percent of respondents were either "very" or "extremely satisfied" using a speech recognition system (SRS) representing an 11 percent increase since a similar study was conducted in 2005.

Nuance Communications commissioned the study involving 262 interviews with customers spread evenly across the telecommunications, banking and finance, and entertainment sectors, concluding that the ability and acceptance of SRS in Australia is maturing and catching up to the rest of the world.

"In the US speech recognition is promoted as a really positive extra service channel that gives convenience, 24 hour access, speed and has always been positioned by the market as something that is very positive. In Australia it has been very different," said Dr Catriona Wallace, director of Callcentres.net.

Wallace pointed to the example of one Australian bank that ran advertisements in the early days of speech recognition touting itself as the bank to call if you want to talk to a real person and not a robot.

"It was a clever marketing strategy but it really positioned the good speech vendors who do have solid and useful applications for customer service in a position where they had to work a lot harder to overcome these original negativities. So Australia is definitely lagging in its uptake of speech compared to the US and Europe, however we're seeing it coming of age now," she said.

Peter Chidiac, ANZ managing director for Nuance Communications, points to advancements in the capabilities of speech recognition systems, particularly open dialogue systems, as one of the reasons behind their burgeoning acceptance.

"The biggest application of SRS that's being deployed right now is what we call open dialogue or call steering application. It allows the caller to take control of the call by openly saying, for example if they are calling Telstra; 'Im moving house and I want to disconnect my line and reconnect it at the new house', and the system understands that.

"We get on average over 80 per cent first pass recognition on open dialogue, but in a directed application where the question is coming from the system and the caller is answering we get up to 98 percent. We have one bank in Australia, a credit union, that has had a system since 2001 and they get between 96-98 percent of transactions completed within the SRS, which is a phenomenal number," he said.

Page Break

Chidiac said that in his decade of experience in the communications industry, the return on investment for speech recognition systems had been on average around 9 months.

"In the traditional IVR world people will press 0 or push a button straight away and not even try to use self service which is detrimental to the company and the consumer hasn't got what they wanted because they then have to wait in a queue.

"The turnaround now is they are willing to use the SRS rather than wait which is a significant thing for an organisaiton. Of course they are saving money by doing that, but the customer is getting what they want which is more important.," he said.

Dr Wallace said that the Callcentres.net study found that customers who were particularly satisfied with the SRS rated it highly chiefly because it was easy to use and understand, fast and problem free.

"Other than the customer speaking to a real life customer service rep, in every other case 59 percent said they would rather use speech than touch tone, 66 percent said they would rather use speech than the Internet, 78 percent would use it over email, 80 percent prefer speech to SMS, and 83 percent would rather use speech than web chat.

"And in fact 32 percent said they would rather use speech than actually speak to a person, so even a third are saying 'give me speech if it works I'll use it every time'," she said.

The survey also gauged how long people were willing to wait on hold before preferring to use an SRS: In 2005 that number stood at an average of 2 minutes. In the latest survey, 57 percent said they would prefer to use an SRS over waiting less than 30 seconds on hold.

On the flipside, 42 out of 262 respondents said they were dissatisfied with their speech recognition experience, predominantly because the system didn't understand their voice and they had to repeat themselves, it was too time consuming, or because the system picked up background noise.