If its performance during a demonstration last week at Microsoft's annual Financial Analyst Meeting (FAM) is any indication, a voice-recognition feature in Windows Vista is not quite ready for prime time.
An interactive voice response (IVR) system in Vista that is supposed to allow a user to dictate text into a Microsoft Word document did not work as expected at the event last Thursday. It failed to correctly recognize what the Microsoft team member was saying on several occasions, the results inspiring laughter from the crowd of analysts and journalists attending the day-long meeting.
When the Microsoft employee told the software to type, "Dear mom," it typed "Dear aunt" instead. When he told the software to "fix aunt," it typed "let's set" instead, and then failed to respond to several prompts of "delete that" in an effort to fix the error. The software experienced several other glitches before the demonstration ended. You can view the demo here.
Matt Rosoff, an analyst with research firm Directions on Microsoft, said he was "surprised" Microsoft would demonstrate the IVR feature of Vista at FAM. "It's not something they made a big deal about, and not something we're following as a big reason to upgrade to Vista," he said. "If it had worked perfectly, it would have been great. Unfortunately, it didn't work out that way."
Rosoff said the feature is the result of new voice-recognition APIs (application programming interfaces) Microsoft is building into Vista that will allow users to dictate instead of type content into Office applications such as Word and PowerPoint.
Microsoft's public relations firm said Monday that the company would not comment on the failed demo.
IVR is just one of a host of enhancements that will be available in Windows Vista, which Microsoft executives said at FAM is still on track to be available to business customers in November, and consumers in January 2007. However, the company seemed to hint that Vista's release could slip again, as Kevin Johnson, co-president of Microsoft's Platforms & Services Division, said at the meeting that the OS will not ship until "it's ready," even if that means it does not meet the current targets for release.
Derek Austin, sales manager Australian and New Zealand for Nuance Communications, makers of speech recognition product Dragon NaturallySpeaking said he did not feel threatened by Microsoft's emergence into the speech recognition world.
"Their entry is growing the market awareness for speech recognition products," Austin said.
He also said any potential threat to Dragon NaturallySpeaking 9 would be to US customers as presently there is no Australian version built into Windows Vista.
Recognising accents is important in speech recognition products. This allows the software to take into account how users pronounce vowels as well as the names of towns and cities (for example, Auckland not Oakland or vice verca).
In Nuance's latest version of its software, Dragon NaturallySpeaking 9, the English language can be set to accommodate English speakers from the US, Britain, Australia (which also applies to New Zealand users) , India and South East Asia.
Released this week in Australia, Dragon NaturallySpeaking 9, which claims 99 per cent accuracy, starts from $199.95.
(Additional reporting by Howard Dahdah.)