Joshua Adelson, Director of Market Development, Brooktrout

NewsBlast Discuss your TR1000 Voice Processing Platform Series. What are the feature sets and how do they impact speech solutions for enterprises?

 

Joshua Adelson Brooktrout's TR1000, which some people describe as a "telephony board," supports voice features as well as speech recognition and fax, and comes in a broad range of configurations, all under a single API. That allows solutions to be scaled up or down, and deployed in just about any telephony environment.  

NB Echo cancellation and audio scrubbing are important considerations when deploying speech, what makes your solution unique in regard to these feature sets?

 

JA This is a critical area you've picked up on. One of the barriers to wider-scale speech deployment adoption has simply been the processing power that's consumed by the speech recognition client software. In a typical system, the client needs to parse out noise, silence and echo from the audio stream before it even starts to recognize the utterances.  We've designed our platform to do audio scrubbing in direct integration with the leading speech recognition engines —a sort of first pass endpointing—before the audio stream even reaches the speech rec client. That way the software only needs to deal with actual speech. We've had this tested independently and it reduces CPU cycles by a third, and system bus traffic by four-fifths.  As far as I know, our TR1000 is the only product of its kind to offer this integration and capability.  

NB Please provide a specific case study discussing an enterprise deployment.

 

JA That's an easy one. We're using it right here at Brooktrout! Working with our partners VoiceGenie and Phonetic Systems, we've created a dial-by-name voice directory. The biggest benefit has been to our mobile employees; they no longer need to program everyone's extension into their cell phone. They just call a single number and request the name. To give you an idea of how popular this has been, we tried taking it out of service so we could use the TR1000 board in our lab. There was a minor revolt! The system was kept online, and we found another TR1000 to use in the lab.   NB How about service providers? Are they using more speech in their offerings?   JA Yes, the trends are common across enterprise and service provider environments. In many cases the service offerings are simply a hosted alternative to an in-house solution. Having the choice between in-house and outsourced solutions is evidence of a growing market that offers multiple choices to its customers.  

NB What do you think can be done to help prospective customers move toward speech enablement?

 

JA It needs be easier to set up and maintain. For our part we've added things like installation wizards that make our boards easier and less error-prone to install and configure. The industry has come a long way, but there's always more that can be done.  

NB How does the adoption of VoiceXML and SALT standards affect board vendors like Brooktrout? Do and should customers even care about VoiceXML and SALT?

 

JA VoiceXML and SALT are good for us because they simplify application development and expand the market. These particular standards also bring tighter integration between speech and web/data applications, which is of tremendous value. There are about 60,000 IVR servers in the world, compared to over 18 million web sites. Speech-enabling even a fraction of the web represents an enormous growth opportunity for the speech industry.   NB What is your definition of 'open systems enabling technology' and why is it important? For years it seemed the world communicated just fine using more proprietary systems.   JA To us, "open systems" means off-the-shelf platforms that run in standard form factors and interfaces, under commercial operating systems. The primary benefit is that the people creating the applications don't have to be bothered with the intricacies of telephony, call control protocols and media format conversion. Imagine if graphics software companies made their own desktop hardware and printers. There would probably be three vendors and the systems would cost $50,000 each. Instead we have dozens of graphics software vendors, good printers are plentiful, and the solutions are inexpensive considering all they do. Actually I think our speech industry has already turned this corner—I don't hear people calling out the need for proprietary systems any more. But the change has been fairly recent so we're only just starting to see the results.