The view from the outside
A useful summary of the state of speech recognition in the enterprise was recently published by eweek (and elsewhere) in an article by Debra D'Agostino of Ziff Davis.
The piece is entitled "Weak Speech Recognition Leaves Customers Cold", but the subtitle is a better guide to its tone and content: Problem: Companies need to take a more realistic approach to speech technologies. The author spoke to a number of companies who have deployed speech systems and to analysts at Yankee and Gartner, and has produced advice to potential customers that is largely practical and sound. Sobering stuff, with some useful stats too. I recommend it to anyone who either thinks that speech recognition is about to take off or thinks that it will never work.
(Yeah, I twitched a bit too at the early implication - including the main title - that core SR technology is to blame for poor user experiences - but this is balanced overall by many of the illustrative problem descriptions, including the Hearst experience of all the work that went into prompting, and the need to bring in "speech experts" to sort things out).
Another aside: everyone I know uses the Star Trek example to illustrate the wild ambitions/expectations of speech technology (or 2001: A Space Odyssey, for a more high-brow/sinister touch, eh, Art? ;-) But haven't we got some better examples yet? I mean, those were years ago - aeons in tech years. Have we no current cultural reference points for the bold directions of this technology? Our symbols are in serious danger of becoming retro-quaint, like early robots with boxy chests and flashing lights...
