TMC Labs
November 2001

 

Speech Enabled Auto Attendant And Media Center

iVoice.com
750 Highway 34
Matawan, NJ 07747
Ph: 732-441-7700
Fx: 732-441-9895
Web: www.ivoice.com

Price: Approximately $2,000 per port

Editors' Choice Award

RATINGS (0-5)
Installation: 5
Documentation: 4.5
Features: 4.5
GUI: 4.5
Overall: A-


Imagine that you own a small business and need someone or something to answer the phone for you as business increases. You may or may not have a legacy system in place, but you’d like the option to add interactive voice recognition (IVR) capabilities. You don’t believe that “IVR is dead,” as some pundits prognosticate. You don’t want someone dedicated full-time to managing the telephony system. You would like something that eliminates the “type in the name on the touchpad” syndrome and “listen to the menu and follow the touchtone prompt” phone menu. 

Is there something that could meet those needs and still be considered “off the shelf”? We had the opportunity to review the Speech Enabled Auto Attendant from iVoice. It appears to meet these criteria.

INSTALLATION
It took us no more than five minutes to set up and use the product before Robert Wiest, systems engineer from iVoice, came by to make sure our Teltone gear (analog phone system emulator) was set up correctly for the demo. 

Today, their preferred system is an Intel-chip-based box (warrantied by IVOX) using a Pentium III 400 MHz chipset, 128 MB RAM, a 1.2 GB hard drive, an ISA slot for the voice processing card (in this case, an Intel/Dialogic board), and either the Windows NT or Windows 2000 operating system. However, you can just add the software package to your existing voice mail system if you already are using Dialogic cards.

DOCUMENTATION
We received a narrow three-ring binder (the 91-page System Guide) with our evaluation unit (version 2.3.1) that did a fairly good job of walking us through the features of this package. The Web site demo/tutorial (available to anyone) filled in the remaining blanks and gray areas. The only thing missing from this particular binder was a glossary of terms, which could have explained such things as “VED” (Voice Energy Detection, a term related to noise cancellation during barge-in) to the layman. Newer operations manuals include a telephony glossary in their appendix covering terms used by iVoice technicians during support calls.

The documentation we received along with the IVOX communications server had a great table of contents. The appendices covered the D/41H and D/21H Intel/Dialogic cards, setup for the Spanish auto-attendant, troubleshooting, and FAQ as well as how to use the name loader import utility with either Microsoft Excel or Access programs.

OPERATIONAL TESTING
This product is intended to be used as a setup-and-forget mode of operation using a Windows NT box provided by ITOX with a Visual Basic-based program that has a speech-enabled application that uses your voice to contact someone within the organization you are calling (assuming you have the “right” accent and they are using this particular system). The system has enough “sense” to try to get an approximation of the name you gave (the auto-attendant makes phoneme suggestions for the human interventionist/programmer) and gives you a choice from a vocabulary of about 3,500 first/last names based either on English, French, German, or Spanish. Adding new ones is very easy. Multiple pronunciations for the same name are also easy to set up in case international business is being conducted.

The speech-enabled function listens to your voice and then phonetically approximates and associates the name with a pre-existing list (the Voice Configurable Dictionary). Think of this as a sophisticated version of Speak & Spell (the toy introduced by Texas Instruments in 1978). Say the name and spell it, and the system will add it to the SQL database telephone directory. A little human intervention may be needed for such things as extension assignments (via the Graphical User Interface Configurable Dictionary). The speech-enabler then can scan through that database next time, when a caller tries to make contact. Callers can also accomplish this if they have access to the right access code. Nicknames can be added to the database if necessary, and “barge-in” technology is also implemented for those who wish to jump ahead in the access-mode approach instead of patiently waiting for the system to finish its premeditated speech processes (don’t worry, it’s quick). 

When we tested the system with a speaker phone setup, it had some difficulty understanding accents, but when we spelled the name, spoke more slowly, and also picked up the handset, it worked much better recognizing the names posted in the database as we tested. Once we got past that minor problem, it responded nicely. We also added names, extensions, divisions, etc., and the system worked as advertised.

One of the disadvantages of using Windows NT as a platform is the issue of memory leakage. For example, while attempting to shut down the system to avoid problems prior to a pending thunder storm, the Windows NT system blue-screened on us. After a couple of reboots after the storm, the keyboard and mouse began functioning again. iVoice is working to accommodate OS changes and upgrades.

The system is designed to also clip onto legacy PBX products as a front-end device. And lest you think you are the first to use this system, iVoice.com has an impressive list of companies and firms that are using their systems posted on their Web site.

This package is intended for small business/home-based businesses or departments within an enterprise where an auto-attendant may make the most sense. However, the nearly $2,000 per-port model (multiples of 2 up through 96) may be a bit steep for many small business budgets, and it currently is locked into a Windows-only configuration. However, if you need a server that does other things besides performing as a communications server, this system is designed to do that. In other words, the voice-recognition engine used isn’t CPU-intensive. The claim is that by using this package, you can save as much as 90 cents on the dollar, as opposed to relying on dedicated full-time customer service representative for each call.

ROOM FOR IMPROVEMENT
Currently iVoice is Windows-centric, but as more businesses migrate to systems such as Linux, iVoice is willing to follow. They want to see the board drivers by Dialogic and others first. They are also investigating VoiceXML applications that can work with their existing software. Web button technology also has captured their interest, and we may yet see an implementation on their Web site.

The Windows 2000 version of their software will be released soon, and beta versions have already been distributed. They are also developing modular code for easier upgrading purposes.

Jerry Mahoney, president and CEO of iVoice said, “iVoice is currently bundling the auto-attendant, voice mail, and unified messaging systems together. The Patient Reminder, Name Dialer, and IVR products are in the process of being integrated with the Media Center applications (auto-attendant, voice mail, and unified messaging), and we are working hard to complete the development of a single bundled package.” Call, check their Web site, or e-mail iVoice for full-package pricing. 

CONCLUSION
Creating a product that works using Visual Basic is relatively easy to do. Making it work well takes special programming expertise. iVoice has captured some of that “special something extra” in their approach to the voice recognition arena.

If you would like something now that eliminates the “type in the name on the touchpad” syndrome and “listen to the menu and follow the touchtone prompt” phone menu within the constraints of the Microsoft computer platform environment as discussed above, then try this stand-alone package.

If you feel comfortable using a package based on Microsoft’s VB, then go for it. Watch closely how iVoice integrates and bundles the various applications discussed above in the near future. 

[ Return To The November 2001 Table Of Contents ]