Speech Technology Begins to Realize Its Potential
Industry experts say speech technology has found its voice.
Specifically, in one of the breakout sessions, called Speech Recognition and the Contemporary Contact Center, various vendor representatives spoke of the practicality of speech recognition in today's contact center. The session, moderated by Shane O'Sullivan, from consulting firm ASA Solutions, included representatives from Automated Responses Interactive, Aspect Communications, First Data, and FutureCom Technologies.
"Why automate telephone interactions?" Jeanne Gokcen, Ph.D., president and founder of FutureCom Technologies, a speech technology provider, asked the crowd. Answering her own rhetorical question she said, "To create satisfied customers and grow your business properly. People want to do business anytime they want to. And, the telephone is still the most preferred way of communication."
Speech technology also offers potential cost savings to those companies that prefer to keep delivering phone-based service. Agent-assisted calls on average range from $1.20 to $4.00 per call, as compared with an average of $0.25 for an automated call, according to Gokcen. This validated the existence of the touchtone IVR industry. But speech recognition vendors promise better results. By enabling customers to say what they want into the phone to get the desired information without bouncing around telephone menus, calls would be shorter and less costly than those using traditional IVRs.
Yet speech technology vendors have had some trouble convincing would-be clients to look past speech technology's stumbling start. "A few years ago we were seeing a lot of companies getting burned by what they thought would be all-encompassing solutions," Katherine Lam, a technology analyst at Datamonitor, told CRM magazine. "Natural-language processing leads customers to believe that our end-users can speak freely in a conversational manner when, in fact, there are still limitations to natural-language technology."
The problems appeared mostly in the systems' inability to understand such verbal vicissitudes as dialect and colloquialisms, as well as background noise. This had a negative effect on customer satisfaction, Lam says. However, new technologies are changing old perceptions. "We're at a point right now where companies have made those investments and have very sophisticated solutions out there," she says.
One of those solutions is directed dialogue, which is the next step below full natural-language processing. It gives people more of a natural way of interacting, without letting people say anything they want, as it only "listens" for keywords. Another approach is directed questions, in which the system only listens for short answers such as yes or no, or a credit card number. These solutions and the further development of natural-language processing technologies are expected to catapult the speech recognition market to $806 million by the end of this year, up 23 percent from last year, Lam predicts. She expects the market to grow between 26 and 35 percent each year for the next four years.
Speech Recognition Technology
...does well with
Yes and no responses
Command and control words and phrases
Smaller grammars and vocabularies
High-quality speech input (microphone)
...has problems with
Understanding letters of the alphabet, especially the e-set (b, c, d, p, etc.)
Larger grammars and vocabularies
Lower quality speech input (telephone)