AdSlot 1 (Leaderboard)

IT Decisions – Voice recognition

Speech recognition technology has come a long way in the past few years. What was once viewed as a helping hand for non-typists and RSI sufferers has morphed into a serious business tool. VARS (voice automated response systems) can reduce operating costs, while generating revenue and keeping customers happy.

Speech recognition platforms, such as those associated with first-generation speech recognition software used on PCs, were largely proprietary. But that’s no longer the case. Today, development is being driven by open technologies such as XML, a data tagging language which can be integrated into most platforms.

Voice XML was designed for businesses wanting to build automated systems to handle basic customer enquiries. It was created by a consortium including Motorola, AT&T, IBM and Lucent Technologies and has come too far for it not to become a de facto standard.

The banking sector has been one of the first to realise the benefits of Voice XML for handling mundane customer enquiries that would normally be handled by counter or call centre staff. Often, such systems have brought about substantial cost-savings. Calls that once cost from £10 to £60 each (depending on complexity) have fallen to £0.06 to £0.50 with voice automation.

Lloyds TSB spent two years developing Phone Bank Express (a 24-hour automated telephone banking service) before it was launched in 1997. Early versions recognised 44 words, but the service now understands whole sentences and handles 1.25 million calls per month.

As in any automation, there are risks with voice recognition. Nick Applegarth, european MD of voice automation systems vendor Nuance, says completely replacing call centres with VARS would be risky. If the human element is taken out there’s no fall back if something goes wrong – and a customer left floundering will most likely go elsewhere. This is why designing such systems is the most crucial part of implementation. “What keeps me awake at night is not the competition but rather one of our customers implementing a system badly,” says Applegarth.

Companies need to be clear about exactly what enquiries they want the system to address. Then, as well as developing databases of vocabularies (recognising different languages, accents and dialects), they have to build in logic to dictate how the system will operate. This process can take months of research and testing. Integration is also key. Although XML is a well-supported standard, it has numerous vocabularies that require transformation before they can interact.

Automated systems must be based on completely reliable hardware, and this demands heavyweight processing, says Mark Blowers, an analyst with Butler Group. “They need to scale up to millions of concurrent users and that sort of reliability is difficult to guarantee,” he says.

Apart from reducing costs, voice-operated systems are simple to use. As a result, Adrian Ewer, european sales manager of mobile information access systems vendor MobileAware, suggests voice technology has a key enterprise role for mobile workers. “Accessing applications on the move is easier via voice,” he says.

Voice portals, which are a relatively inexpensive way of making content accessible, are one new concept. By providing information for phone users, they could generate huge revenues. “For years the issue about voice has been – does it work? But that’s history. It’s now a question of finding the right application,” says Applegarth.

Related reading