/ 25 December 2003

The future finds its voice

Checking flight arrival and departure times or the status of your bank account has never been easier. With the advent of speech-recognition software, and its more advanced application, you are now able to avoid the call-centre loop and get the information you require by speaking directly to the telephone.

If you have made a call to the Airports Company South Africa (Acsa) recently you would have spoken to the automatic voice prompter, which takes you through the process of finding out flight times and numbers. Any call centre is supposedly able to deliver this kind of service, but in reality varying technologies provide varying levels of speech-recognition intelligence.

So how does it work? According to Mike Renzon, MD of Intelleca Voice and Mobile, new speech-recognition systems are different to the traditional Interactive Voice Recognition (IVR) systems we were used to in the past. IVR is able to drill down into “specified” data to try to derive the right information, but they are still limited in their ability to intelligently prompt or farm information.

Newer speech-recognition software lives above your Web server platform and gathers information from various sources, not only those on your own database.

“What people don’t realise when they are speaking to that electronic voice on the other [end of the] line is that you are actually speaking to the Web server, and ultimately the Internet. We build the applications logic around what is sitting on the Web server through the use of technology called Voice XML.

“Exactly like when you open a Web browser like Netscape or Explorer the system gathers the information in the background, and instead of marking it up in text or HTML it marks it up as voice,” explains Renzon.

But where does the real application for this technology lie?

When a speech-recognition application is deployed it can automate anywhere between 60% and 70% of the traditional processes of a call centre, Renzon says.

It is all about streamlining processes. Companies have to start looking for smarter ways to deliver their services and simultaneously reduce costs. Speech-recognition software can reduce the amount of time customers spend in the phone to a call centre, as well as reduce the actual overhead and number of agents used in a call centre.

But with any technology-driven service there are barriers to entry. Renzon says speech-recognition companies have to work hard to ensure that the voice-activated response on the other end of the line is geared to offering clear instructions and is as personal and as realistic as possible.

He adds that the only way people are going to become comfortable with the technology is through constant use. There is no training substitute for exposure.

“There are a number of applications for both speech- and voice-recognition systems. Financial institutions can make use of speaker biometrics or voice recognition for authenti- cation, ultimately ensuring heightened security. The system will record the utterances of each conversation and feed these back to the system. The technology is now at a point where it no longer requires you to spell things out, it can be loaded to read inflections and accents in dialects,” says Renzon.

In the case of the Acsa application a caller is taken through a series of short questions to establish what they require. If the system is unsure of the user’s requirements it prompts a new set of questions, drilling down further into the application to extrapolate data.

For example, if you are unsure of the flight number, but know the airline, the system gives you the opportunity to select either option through a simple series of yes or no answers. I called the system to test its limitations and found out which flights were scheduled into and out of Kenya. The system is designed so as not to complicate the process and there is very little detail in acquiring the information, the detail comes later with the delivery of the information.

This speech-recognition system is a work in progress and companies using it should work closely with customers by reviewing the database of calls to see where the users feel uncomfortable. This can be done through analysing hesitation and possible stumbling blocks from the user in the delivery of service and then tweaking the system to ensure complete ease of use.

“Slowly but surely the technology is filtering into the market … The easy objective for clients is that it is a lot less expensive for contact centres to use for straightforward queries and issues as it eliminates the need for huge capacity in the contact centre. Speech-recognition software can facilitate the fact that queue times go right down and that people who have basic enquiries can be dealt with by the automated system reducing their own frustration and better facilitating their experience with your company,” adds Renzon.

The technology is not new, neither is the experience, but the application and intelligence of the system has come a long way. In fact, the question nowadays should be more around why don’t you have speech-recognition software as opposed to why do you have it.