Voice Access Technology

 

I have a dream.

A dream that one day, we will wear computer technology and it will help us, guide us, entertain us, anywhere, anytime.

Wearable Computing

Most wearable computing systems are based on some type of processor board that is worn in a pouch and some type of "heads-up" display technology mounted in front of your eye.  But voice technology is cheaper and easier and more portable and safer to use when driving, walking, etc.   And to get access to a lot of information, you really need wireless internet access.   At Voice Access Technology, I designed and built a system to provide internet and other information access via a wearable wireless voice-based device (to wit: a cell phone!).

Voice-Based Applications

The system's centerpiece is a VXML interpreter that I designed and built which uses Nuance speech technology to run custom applications.   Telephony is routed from the PSTN (public switched telephone network) to the voice servers through VOIP (voice over IP) using a Cisco 53xx series gateway.

To further enhance the flexibility and robustness of the system, I designed and built a VOIP (SIP is the voice protocol used, actually) "re-director" which maps dialed phone numbers (DNIS) to specific server groups.   It even balances the load between servers in a group and generates notifications when a server has a problem.

VXML - Voice XML

I'm most proud of the VXML interpreter, though.   Initially, I wrote an interpreter which used XML DOM to parse the VXML and then built a parallel internal parse tree.  Since VXML can include embedded JavaScript, I used the Microsoft Scripting control and maintained a synchronized "script context" with the internal context.  It worked, but it wasn't pretty.

The World's Only VXML Compiler

VICAR was my attempt to create a VXML compiler.   It stands for "VXML Interpreting Compiler And Runtime".   In this system an XSLT stylesheet is applied to the VXML and converts the VXML tags and the embedded JavaScript into a single (larger) JavaScript.   The generated code references some external objects for its voice functionality, but basically it maintains all its own state and runs autonomously!  This eliminates all the messy "synchronizing code" of the previous version and decouples the application from the Nuance system -- another recognition system could substitute just by supplying a small number of functions.

Voice Product Registration

Voice Access Technology built a set of VXML application modules (like name and address and birthdate, etc.) and pursued contracts for  "Voice Product Registration".   You know those little registration postcards that (almost) everyone throws away?   Now those can be filled in by telephone.   And most people don't even have to say much, because the "reverse lookup" databases look up the caller's phone number (ANI) and just prompt for verification of that name and address.

Customers

One customer, Holmes Products, signed up for voice registration for their humidifier and air purifier product lines.   After a year of no additional customers and reduced call volumes as the product stickers sold through, Voice Access was done.

The End

ŠJason S. Loveman, 2003
Last modified: August 1, 2003