Market Leaders: Speech Engine
After 2009’s high-profile acquisitions and panic in some sectors with the full and undeniable onset of the recession, 2010—though not as boisterous—is proving to be an important watershed for speech engines. The market for speech engines continues to grow at a steady rate as more applications are rolled out and with elements that rely more on network-based processing. Speech engines are creeping their way into mass use beyond the traditional base in customer service, particularly for everyday consumers on mobile devices. It’s beginning to look like the bad old days—when inaccurate speech recognition and heavy computer-voiced text-to-speech tarnished the industry’s reputation with the masses—are finally beginning to recede.
The vendors in this year’s speech engine award category represent some of the best and the brightest that have helped to reverse perceptions.
Joining the market leaders for the first time this year is AT&T. Though it has been a rather sleepy news year for the company’s speech technology research lab, the firm was able to rise up this year because of its decently rated product and—maybe more significant—the space left vacant by perennial leader IBM, which dropped from the race with the licensing of its speech portfolio to Nuance Communications last year. AT&T scored particularly well in affordability and customer satisfaction, with a 4 in both on a five-point scale. The company also enjoyed an adequate rating in accuracy with a 3.5.
A consistent leader in the speech engine category—having garnered the top spot in 2007, 2008, and 2009—Loquendo was just barely inched out from claiming a repeat win this year. So fierce was the competition that the firm was beaten by less than one-tenth of a point. Loquendo achieved high marks across the board, with its strongest showings in customer satisfaction, accuracy, and company innovation. Innovation, in particular, has been a pretty solid source of company pride. Just this past April, it launched its 70th TTS voice, making its stable of voices the largest in the industry and sprawling its offerings out across 28 languages. If the company continues to grow, then expect to see it return to the winner’s seat.
Rising from Vendor Contender last year, Microsoft/Tellme is this year’s big winner. With a perfect 5.0 in affordability, a 4.0 in customer satisfaction, and decent scores in company innovation and accuracy, Tellme was just barely able to muscle out this year’s competition by the slimmest of margins. The source of much of that muscle this year is its hosted delivery model, which has allowed the company to make significant gains in the accuracy of its product. “Because they’re hosted, they’ve got this database of millions and millions of utterances, and on a project-by-project basis they really make use of it,” explains Susan Hura, principal and founder of SpeechUsability. “You’re going into any given project a lot more informed.”
Expect the company to continue getting stronger, too. Keeping an eye out for contexts beyond the enterprise call center, Microsoft/Tellme is pushing hard in the automotive and mobile spaces, capturing even more utterances from a multiplicity of scenarios.
Also worth noting was Nuance Communications’ engine, which was the clear winner among all the vendors this year as far as accuracy ratings go. Our judges consistently thought its engine was one of the strongest pieces of technology out there, which shouldn’t come as a surprise. Nuance has consistently ranked as a Market Leader, and its accuracy has often been singled out as a major strength. In fact, this year it posted a perfect score of 5. However, the company’s steep pricing scheme—long a source of complaint—left it with only a 1.3 score in affordability, tarnishing its overall rating and holding it back from placing in the top three for the first time in the past four years.
“They are very expensive for what you get. I hear it from client after client: Nuance is hammering me,” says one consultant.
“For English and the dominant languages around North America, there’s no one better,” said another, adding that even Microsoft is still not up to where Nuance is in accuracy.