::scr on agents[0] was on bots

[jo walsh]

> It would be fairly easy to hook up a speech<->text<->speech pipeline to
> the bot, I wrote a quick irc->speech bot when I was 'Resting' last year
> and putting in something like Sphinx wouldn't be difficult.
> Kevin Lenzo has already done it at least once.

i had a go last week, but though i got banal but encouraging results from
sphinx-demo, i got horribly lost in the documentation. mostly i was
messing around with the response/ text-to-speech part, making little
newsreader / music-mixer modules &c. (the rab voice in festival is really
nice btw, a very english sort of synthetic voice).a speech input interface
would be cute but for my applications it would be only one of many
interfaces, whichever interface was closer to you according to what you
were doing or how you were feeling.

i guess a speech output interface is nice especially for me because it is
easy to multitask through; just switching on cognition partially and
letting the brain check for highpoints and patterns, like the news on the
radio. but speech is not fast you need to focus and typing may be quicker
a lot of the time. otoh typing is quicker than mousing all of the time, cf
all these mac users' keyboard shortcuts, and an element of the interface
isnt integral / in some given context irreplaceably useful, then it's
just going to be switched off.

i guess a speech/speech rec  based interface to the machine/the net  would
be a very approachable alternative for some people specially where
literacy or motor skills were a question, but the toolz would have to
completely transcend what we have now in festival and sphinx for that to
work. and it could only be one of a clear set of alternatives i think.  

z

[0] i do have an awful lot of wibble backed up about the whole agent/bot
question having got some good feedback off (void) but am trying to write
it all down proper with structure like and a url.