Gates on speech recognition: a long road still to travel
Needs the whole Windows trip before it works, natch
MS Analysts Speech recognition software, and to a certain extent handwriting recognition and machine translation, have been a sore point at Microsoft for some time. An enormous amount has been spent on research in these areas, but still there is no sign of a product.
Speaking to financial analysts recently about .NET, Gates said that "Speech, handwriting, even sort of visual gestures, will be part of that interface... The new interface doesn't require that you totally work through speech. The keyboard is still there. And, in fact, in the early years that will be the primary way that people get at it.. It's not a simple thing to do. We can't just have one speech application or one handwriting application; we have to do exactly what we did with Windows, and get that base out there, get the tools out there, and get thousands and thousands from people who are used to working with the platform at that level."
Gates said he expected speech input to catch on in Japan and China because of keyboarding complexity. He commented that there had been a lot R&D in the entire voice area, with "hooks that we put into Windows itself so that applications can interact with voice applications."
"It's a non-trivial thing," he said of this product area where Microsoft had failed multiple times to produce a product. Gates ventured the opinion that handwriting recognition would be mainstream before speech recognition - "in two or three years time", so a Microsoft speech product looks like being perhaps five years away, which could mean never, as it is most unlikely that it could overtake Dragon/L&H, IBM or Philips. ®