Good business requires that employees and customers both have easy access to information. Implementing today's Internet-based tools and devices, however, isn't always the most effective or convenient way to accomplish the task. To better assure access to your company's vital data, IBM has enhanced the WebSphere e-business platform with the release of WebSphere Voice Server 1.5, dramatically improving the thin-client capabilities of a standard telephone.
Voice Server nestles between your VOIP (voice over IP) gateway and application servers to allow telephone users to access enterprise applications and data resources using both voice input and a touch-tone keypad. Although IVR (interactive voice response) systems are not new, Voice Server provides a means of voice-enabling your existing Web infrastructure, allowing voice applications to be built on top of the same data repositories and business logic used for existing channels.
Voice Server 1.5 lacks some features available in emerging voice products, such as the drag-and-drop IDE (integrated development environment) for VoiceXML applications found in HeyAnita FreeSpeech. Also, Voice Server currently lacks out-of-the-box applications to jump-start your efforts.
Nevertheless, in our tests Voice Server proved to be robust and capable, and the included ViaVoice engines for speech recognition and text-to-speech output were superb. Taken as part of a total WebSphere Web infrastructure and management solution, WebSphere Voice Server 1.5 offers a good means for deploying Web-based voice response applications with significant potential for improving information accessibility.
Of course, routing calls from your PSTN (public switched telephone network) to the Voice Server on your IP network will require installation of an appropriate VOIP gateway. Voice Server supports ISDN PRI (Primary Rate Interface) and CAS (Channel Associated Signaling) connections, and it contains an H.323-compliant interface to manage calls to and from the gateway.
We found installing and configuring Voice Server to be very straightforward. However, integrating Voice Server into your VOIP network will require the assistance of telephony administrators who will need to address routing, load balancing, and connectivity between the gateway and Voice Server.
Incoming phone calls are routed through the gateway to the Voice Server, where they are handled by Voice Server's VoiceXML browsers. Each incoming call must be answered by a separate browser, so you'll need to experiment to determine how many browsers to run to avoid greeting callers with a busy signal. The gateway can be configured to help balance the load.
The browser presents application menus and dialogs to the caller, responding to his or her voice commands and touch-tone input. Voice Server employs the Speech for Java interface, based on IBM's ViaVoice technology, which deciphers the caller's input and renders the VoiceXML documents that drive the caller's session.
Log files maintain records of server condition, alert conditions, and usage. Because server performance can be adversely affected by a number of factors -- everything from the complexity of your VoiceXML application to a glut of incoming calls -- we would like to have seen more intuitive ways to avert error conditions before they arise.
However, we liked the control we had over server activities. For example, we could configure the server to accept all incoming calls, accept only transferred calls, or accept calls only from certain DNIS (Dialed Number Identification Service) numbers. In total, we found that Voice Server excelled at voice recognition, particularly on partial phrase matches, and Voice Server performed reliably even when we interrupted its responses with unprompted voice and touch-tone input.
IBM offers a free SDK (software development kit) that provides all the tools you need to begin creating and testing Web-based voice applications. The SDK includes the speech engines, a VoiceXML browser, a DTMF (dual-tone multifrequency) simulator for touch-tone input, and a servlet for constructing the grammar files used by the browser.
Although VoiceXML doesn't require special tools for development, any IT shop would benefit from testing code validity prior to deployment. Absent in the current Voice Server offering, VoiceXML development tools can be found in WebSphere Studio, which interfaces with the SDK's VoiceXML browser to offer wizard-driven coding assistance.
Despite the restrictive deployment requirements of Windows NT and a Cisco 2600 router, we found that IBM WebSphere Voice Server 1.5 represents a good means of offering new services to customers and employees. With the current ineffectualness of mobile browsing and delays in the rollout of 3G (third-generation) wireless, Voice Server's modest price and minimal learning requirements make it a worthy alternative to consider.
Managing Analyst James R. Borck (email@example.com) covers enterprise e-business solutions for the InfoWorld Test Center.
THE BOTTOM LINE: CONSIDER
IBM WebSphere Voice Server 1.5
Business Case: In addition to providing more flexible data access to road warriors, this IVR system can reduce customer support costs via customer self-service.
Technology Case: Voice Server allows you to build voice applications atop existing data sources and business logic, easing development and deployment efforts.
-- Good voice recognition and text-to-speech capabilities.
-- Robust server response.
-- Call forwarding.
-- Supports English, French, and German.
-- Limited deployment environment.
-- Limited samples and prebuilt applications.
-- No support for developers.
Cost: US$15,000 per CPU.
Platform(s): Windows NT 4.0 with Service Pack 6a; requires Cisco 2600 router with IOS Version 12.1.5.
Company: IBM Corp.; www.ibm.com.