Aumtech, with help from Microsoft, breaks through speech rec cost barrier.

Automated speech recognition (ASR) or speech rec, where people converse with computers, has long been a 'tomorrow' technology. ASR's promise lies in its ability to provide near-contact center agent-quality customer care for basic interactions like obtaining flight information and order taking at a fraction of the cost: 50 cents per transaction compared with $5-$9 with live agents.

ASR, because it is conversational, is also much more end-user-friendly and versatile than keying in numbers and letters on phone keypads i.e. DTMF or Touch-Tone (TM). That means more people will use automated systems instead of or before they 'zero out' to agents. The applications have developed sufficiently and are in enough widespread use to gain acceptance by the public to the point where it has been 'trained' to speak to the computers.

Yet for all these benefits and feasibility ASR has been deployed in only 25 percent of contact centers. The barrier has been steep costs: up to $3,000 for a single software license, which translates to $150,000 for a 50-seat contact center or office, plus application development, installation and maintenance fees.

That barrier has now been smashed.

Aumtech, which makes IVR platforms, has, under contract and in collaboration with Microsoft, has broken through the cost barrier by developing and recently successfully deploying a unique software interface, written to industry standards, linking Microsoft's ASR tools to the IVRs.

This innovative tool and partnership has knocked ASR costs out of the sky, plummeting to as low as $10 per license, and dropping the total cost of ownership by 75 percent.

Aumtech and Microsoft revealed the details during a webinar hosted via Technology Marketing Corporation's site, TMCnet.com, Thursday Aug. 14, on the eve of Speech TEK, a trade show covering speech technologies that took place Aug. 18-20 in New York City.

The interface, the MRCP Connector, enables Aumtech's customers to use Microsoft Speech Recognition Services even if they do not have Microsoft platforms. MRCP is the industry standard for linking ASR engines with IVRs. Microsoft's Speech products do not have a built-in MRCP interface.

Aumtech is the only supplier of an MRCP-based solution that connects into Microsoft's ASR applications. The Aumtech platform is carrier-grade and works with VoIP and with TDM/PSTN phone networks.

While Aumtech also supports ASR products from other leading vendors such as Loquendo and Nuance, it is with the Microsoft solution where the savings have been achieved.

[ILLUSTRATION OMITTED]

Microsoft's leveraged its speech applications that are built into Office and Vista to offer a low-cost packaged solution that requires much less systems integration work, which its competitors charge extra for, according to Aumtech.

Microsoft's solutions are built on the Sphinx and HDK speech engines pioneered by Carnegie Mellon and Cambridge University respectively. They support both directed dialogue and natural language, which it calls Conversational Understanding', and in a small but growing list of languages.

JetBlue is one of the first customers and the first major corporation to have gone live with the Aumtech/Microsoft solution. JetBlue uses ASR to supply flight information: a strictly scripted, frequently tapped and customer-critical application that is ideally suited for this technology. The airline presented its case story for the first time on the TMCNet.com webinar.

The firm's CIO, Duffy Mees, went out on a wing with it in early 2007, which flew later that year, and which has since provided information to callers with amazing accuracy and few zero outs.

The Aumtech/Microsoft product replaced an application sold by Philips in 2001 and deployed on an Aumtech IVR. Philips' ASR division was later acquired by ScanSoft, which renamed itself Nuance when it bought the firm of that name.

The Philips/Nuance ASR application had been reasonably successful. It achieved an over 88 percent accuracy rate, a 60 percent call completion rate, and an average 91 second call duration that together saved the airline money compared to having the inquiries answered by live agents.

JetBlue needed more, however. The current system could only handle 46,000 calls per day and the airline, and inquiries, were growing fast. Yet when the air carrier, which prides itself on high quality service at low costs, saw the price of a license upgrade, it became stricken with Sticker shock: $175,000.

Duffy Mees then approached Aumtech, which had been working with Microsoft on the MRCP Connector. Aumtech came back with a solution that could be delivered for less than $47,500.

The new solution took flight in late 2007, and JetBlue could not be happier. Call completion rates rose to 82 percent, while recognition accuracy rose above 90 percent. Call duration shrank to under 60 seconds. The new system can now handle over 127,000 calls per day. Cutover was seamless; both solutions ran in parallel until the new one was off the ground, and calls were ported over with little or no dropouts.

"The Microsoft engine, connected to our Aumtech IVR, offered equal if not better quality ASR than our previous solution but at a much greater value for money than that product's newest model, and was far easier and quicker to install," explains Mees. "We could not be more pleased."

So is Aumtech. It has four other clients already under contract. Immediately before and during the webinar it received several leads from potential new clients. Several IVR vendors are in discussions to provide the Aumtech Connector/Microsoft Speech Server combination to their own client base.

"With performance, accuracy, and tools to migrate existing applications, and with Jetblue having proven that high performance ASR licenses can be delivered on an Aumtech platform at costs 1/50th of the price of competing solutions, how can the rest of the industry maintain its premium price strategy?" asks Aumtech COO Tom Porter. "Not long, because existing and prospective customers are realizing that they no longer need to pay those high fees."

by Brendan B. Read,

Senior Contributing Editor, Customer Interaction Solutions