[alicebot-archcomm] 3D FACEs

[Kim Sullivan]

> The AIML spec should define how different
> systems like facial sythesis can be access via a consistant interface and
> how the client can inform the server of its requirements & capibilities.
> The reference model (Alice) would then implement these interfaces and the
> drivers for a selection of facial experession engines, to demonstrate how
> to build new drivers (integrate a new engine).

  I agree with the idea of a constant interface for clients, but I don't
think that modifying the AIML spec itself to incorporate facial expressions
is a good idea.

  What I would like AIML and alicebot to be, is a thin natural language
interface between the user, and the computer. This basically means that AIML
handles only the stimulus-response part (returing a specific template for a
given input), keeping all other processing (face, voice I/O,
command&control) to the 'client software'. By 'client software' I mean the
program that the user uses to type in (speak) his inputs.

> Embedded commands:
> The commands could be embedded into the aiml template. A whole new set of
> tags would be required for systems that require facial expressions, but
> would be stripped out before sending the text over the wire (for clients
> that did not have facial support. Tonal inflections is another area that
> has come up, that has the same kind of temporal issues.

  The mechanism is correct (embedding the additional information into the
template), allthough I personally think that the issue of tagset & filtering
should be wholly left to each individual client/server implementation,
without including it into the AIML specification (at least not now).

  Let people first make such a system (they can even use Program D as a base
if they want). Let the developers resolve all issues (tagset, protocol, how
to register different drivers, how to query capabilities...) and then show
the comitee the working results. I'm afraid the opposite approach (first let
the comitee work it out, then implement it) has very rarely worked out,
often resulting in quite heated arguments (flame wars), and then the issue
is forgotten anyways.

  Please don't take me wrong, I'd -love- to see an alicebot with an animated
face (if it had working speech reco and synth & command/control caps I'd be
in heaven), but I'm personally not prepared to discuss this issue on the
archcomm. Use the regular list, or the developers list.

Best regards,
Kim