On 7/11/07, Ovidiu Sabou <ovidiu.sabou(a)gmail.com> wrote:
and technology isn't yet ready(speech recognition has more to go) for
talking to computers the same way as talking to humans, but we can already
develop software, based on existing speech recognition engines, that
should
be usable in a few years after the speech technologies mature.
Don't get carried away by the progress. There is an in-principle reason why
speech recognition will (probably) never work. In particular: to resolve
sound and meaning ambiguities, you need to be able to understand humans. To
understand them, you need to know about how humans are, how do they react,
what their intentions are, how do they feel, what they possibly mean or
want, what are their cultural practices and so one.. Sure, you can make a
toy system that could respond to some fixed questions in some fixed-pitch
voice and accent or such, but I would guess it would be more frustrating
having a ignorant avatar, than having a normal edit/search/text box.
Robert