Re: [Communicator-user] Communicator and VoiceXML?

[Dave S. <sta...@bb...>]

"Bradford W. Miller" wrote:
> 
> Turing machines, however, are a bear to program. What's far more important
> than simple theoretical computational power is having the appropriate
> expressive power, i.e., the right abstraction. Attempting to impose the
> right abstraction on a state machine is probably not the most efficient use
> of resources, just as I would not want to try to write the X window system
> in PROLOG (which is also Turing equivalent to more or less everything else),
> let alone use it once I was done.
> 
> Having the right architecture and abstraction makes it easier to envision
> and create the right solutions. I think the community is still searching for
> the right thing, but (in solid agreement with Alex (for a change! :-)) I'm
> pretty sure it isn't state machines.

Exactly.  In *theory*, you could develop a Communicator system using a
state machine formalism, but in practice the number of states and links
becomes prohibitive.  MIT had a paper at an ACL Conversational Systems
workshop a couple of years ago that illustrated this quite well.  We,
like CMU, started out early on with a state network, but quickly gave it
up in favor of a more flexible control scheme, as did MIT and U.
Colorado. 

I agree the community is still searching.  I think a standardized dialog
management formalism for Communicator-level systems is years away and at
this stage of development, that's fine.   Everyone that I know designed
their own, and (of course) thinks that theirs is the best.  I know I do
;)

   Dave