What three tips would you give a company who is interested in Speech Recognition?
Companies can implement successful speech applications by following the following three principles:

• First of all, educate yourself about the advantages and, most importantly, the limitations of speech technology, which requires a fundamentally different design approach from traditional text-based applications.

• Second, target your speech applications at a specific problem domain. Remember that each speech application needs to have a very unique set of requirements for Voice User Interface (VUI) and grammar design and requires extensive usability testing.

• Finally, it is critical that you strike the right balance in your Voice User Interface design. Ideally the system will have a balance between the simplicity of touch-tone IVR systems and the complexity of AI-like speech applications. While IVR systems are universally disliked, AI-like speech applications are not practical with the speech technology commercially available today. Including elements of both will provide the best solution.

What sort of mistakes do you see other companies make when dealing with Speech Recognition?
Speech technology has made tremendous progress in the past few years in both ASR and TTS, but we are not quite ready for StarTrek-like speech applications. I believe the biggest mistake that companies make while dealing with speech applications is setting unrealistic expectations in terms of the complexity of dialogs that a caller can have with the system. As a result, the benefits of speech applications have been often oversold.

How do you think we will be looking at Speech Recognition in five years time?
In five years I envision ubiquitous deployment of multi-modal applications running on wireless handheld devices with speech interfaces acting as a primary channel for data input. Also, improvements in TTS technologies will result in much more Web content available in audio format. In general, I believe that all user-computer interactions, even on the desktop, will feature more natural and easy to use interfaces driven by speech.

What is the biggest application for Speech Recognition?
In my opinion the biggest opportunity for speech technology today is in call centers. The so-called “call deflection” strategy, under the motto “do not call us, go to our web site,” does not always work as well as expected. The lion’s share of all customer support inquiries (over 80%) is still handled over the phone. So, there is a tremendous opportunity to deploy speech applications that reduce costs and improve customer satisfaction, while providing very compelling ROI. At the same time, companies investing in speech applications today have a unique opportunity to prepare for the emergence of converged communication networks which bring data and voice services together. To do that, companies must consider implementing multi-channel data-access strategies that enable access to corporate data and customer-facing knowledge bases using speech recognition, Web and handheld devices.

What are some barriers we face when deploying Speech Recognition?
One of the biggest barriers is the lack of off-the-shelf speech applications provided by Independent Software Vendors (ISVs). Today’s speech applications are typically custom built by consulting companies or call center integrators. From that perspective, I am very excited about Microsoft’s recent announcements on SALT and .Net Speech SDK, products that are targeted squarely at the software development community. My other concern is an insufficient support for speech standards, which drives the continued proliferation of proprietary speech and IVR systems. Yes, VoiceXML is here today and many IVR vendors continue to pay lip service to it, while, at the same time, pushing their own proprietary systems, which are complex and costly to implement, maintain and upgrade.

About the Author
Zor is a founder and CEO of Tell-Eureka Corporation. He was the founder and CEO of BuzzCompany.com, a premier developer of collaboration and messaging software, which was acquired by Multex.com (MLTX) in 2000 where he served as Senior Vice President. Before that, Zor was a Senior Consultant at Microsoft Corporation, and held management positions at both Computron Software and Information Builders.. His interest in speech technology dates back to 1992 when he worked at Bell Labs.

About Tell-Eureka
Tell-Eureka is a pioneer in the development of multi-channel self-service customer support applications that combine speech recognition and Web technologies to cost-effectively automate first-level customer support, while leveraging existing investments in online support initiatives. The company's intelligent self-service solution enables callers to resolve common problems quickly and effectively using natural conversations over the phone, freeing customer service representatives to focus on more complex problems and revenue-generating activities. For a demonstration or more information, call 646.792.2718 or visit www.telleureka.com