Microsoft Begins Broader Test Of Speech Software

Speech Server will compete with VoiceXML standard, which is used by several other vendors.
Microsoft on Wednesday began its first beta test of computerized speech software designed for call centers, paving the way for a server product expected next year. The pre-release software, called Microsoft Speech Server, uses a company-backed standard for application development that's starting to find adherents. But Microsoft's approach will face tough competition from an IBM-backed approach that's being used by companies including American Airlines and Merrill Lynch.

The first beta test of Microsoft Speech Server could involve as many as 100 companies, up from the 11 that have been testing a "technical preview" released last fall, according to Kai-Fu Lee, Microsoft's VP for speech technologies. Microsoft won't say who those 11 are, but says more than 40 resellers, independent software vendors, and original equipment manufacturers, including Hewlett-Packard, Fujitsu Consulting, and Scansoft, have undergone training on the new software, which will require Windows Server 2003.

Microsoft's software--a server and development tools that entered a third round of beta testing on Wednesday--includes the ability to write apps that recognize spoken commands, convert text to speech, and generate spoken prompts in a menu. The software is due in the first half of next year. It's designed as an alternative to dedicated systems used in companies' call center operations, and to apps that use VoiceXML, a standard backed by IBM, AT&T, Lucent Technologies, and Motorola. Companies including Amtrak. American Airlines, Merrill Lynch, Sprint, and United Airlines use VoiceXML. Microsoft's proposed standard, Speech Application Language Tags (Salt) is designed for use by Visual Studio developers.

"Microsoft is coming into the market with their own competing platform and their own protocol," says Elizabeth Herrell, an analyst at Forrester Research. "It's going to be a race of which developers support which protocol rather than what the end-users see on their desktop."

Companies increasingly are choosing to write voice-response apps for navigating corporate phone directories and completing customer service transactions using industry standards such as Java and .Net, instead of systems that rely on custom code, which can cost more to maintain, according to Herrell. Microsoft's initial release will likely work best for relatively simple applications such as locating employees in a phone directory, rather than for more complex transactional apps, she says. "This is only version 1.0."

But according to Lee, Microsoft's integration of its speech-development tools with its widely used Visual Studio tools suite could provide an advantage. "The tools make Salt even more usable," he says. And as faster, data-capable wireless networks proliferate in coming years, Lee says Salt will make it easier to write apps that combine speech input and output with graphical user interfaces.