Re: HumanML and VoiceXML etc

[JohnRobert Gardner <John.robert.gardner@sun.com>]

A few thoughts on applications and sources in this thread:

Manos Batsis wrote:
>
> I suggest we start discussion on future directions. My quarter of an
> euro:
> 
> 1) Study current non XML popular classification systems (from DDC[1] to
> NAICS[2] to "Classification of Living Things"[4] (interesting, includes

The research references lying behind books such as "The Naked Face" is a
good place to look for categories of non-verbal expression (plus a lot
of speculative--not relevant--material on what different facial features
indicate of character . . . ), as well as the classics such as
Bridwhistle's earlier work on classifications of non-verbals.  With
special thanks to my old non-verbal communication mentor, Joe
MacDoniels, a few tidbits noted here.

Birdwhistle worked in a sort of heirarchicall OO fashion, following
linguistic theory of the time.  He broke nonverbals down to individual
components (his term for this escapes me), that combined to form
descrete messages, combining in turn to complete non-verbal phrases.  He
abandoned this "semantic" interpretation later, feeling instead that the
nonverbal cues were not mng. or semantic content, but rather containers
that were parts of a complex of mng generators, and the nonverbals were
but one ifx for it.  In short, his classification scheme, with his later
relational reading on it, would fit _very_ well as a starting point for
thinking about a straw man.  His later work looks at the classifications
as syntactic pointers . . . less evaluative, I think, for our purposes
and hence more useful than would be wading into semantic interp.

In addition, in the group space (did we consider that there are human
inflections that exist and are visible only in the group space?  we
should consider this area for markup as well . . .) is Aubrey Fisher
looking at relational dynamics--syntax--of group coding.

> humans) to whatever we would like to focus on by building modules). This
> would provide us with a jumpstart on any topic we might want to get our
> hands in. It will also help transitioning existing knowledge bases.
> 

Agree, and to this issue, cf. point on applications of HML below:

Manos Batsis wrote:
> 

> What we actually need in short is progress leading to real applications.
> 

Following Manos' point here and also the following from Steven: 

<snip>
s.livingstone@btinternet.com wrote:
> 
> Has anyone had any thoughts on how this could potentially work with VoiceXML.
> 
> What kind of technologies are availabel that coudl link these together? For example, someone phoning your company and some "mood detection" software combining VoiceXML and HumanML to determine priority and call types? I imagine it is pretty easy to determine "angry" and "relaxed", but not sure what else? I guess keywords used by the caller could be used to determine the mood.
> 
> I don't know much about the "mood detection" software and in some cases it so
</snip>

I think HumanML in this kind of VoiceML space is better conceived first
in the space of server-side delivery.  IN other words, whether it's a
phone-menu-tree or a voice internet gig, or some Java/Jini gizmo with
voice recog. and response, as soon as we're thinking of CMC, the
realization of how auto-voice feedback can have tailored tagged
inflection options (e.g., the angry caller impatiently pushes item 1+* 8
times, the voice generating software could select the phrase "I'm sorry
for the delay, but repeated pressing will <inflect
type="gentle-admonish">not</inflect> expedite." (ignore the semantics,
just a hypothetical notion) - there could be one storage of that phrase,
with--for instance--"not" marked up with gradient options of strong or
weak admonishment.

Similarly, in the area of storing human communication, this is a key
area to record such details.  It seems a more reasonable, and ready,
business case and ROI lies on uses of HML with _existing_ human
communication that must be delivered/stored than to start first with
"recognizing" inflection.  The programming for hte latter is too
subjective.

just a couple pennies worth . . .

jr 
=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-|
John Robert Gardner			|
Deconstruction Architect		| To live is			
Technology Office, IT			| The rarest thing			
Sun Microsystems Inc., UBUR02-306	| In the world:
1 Network Drive				|
Burlington, MA  01803-0903		| Most people exist		
Ph. 781-730-4012/#6-9482 fax 442-1539	| That is all
e-mail  john.robert.gardner@sun.com	|
----------------------------------------|
http://vedavid.org/diss/
OPINIONS AND VIEWS NOT NECESSARILY THOSE OF SUN MICROSYSTEMS, ET AL