W3C home > Mailing lists > Public > public-xg-emotion@w3.org > November 2006

Re: [EMOXG] Requirements UseCase 1+2+3

From: Marc Schröder <schroed@dfki.de>
Date: Thu, 30 Nov 2006 17:23:54 +0100
Message-ID: <456F059A.4050103@dfki.de>
To: EMOXG-public <public-xg-emotion@w3.org>


I have not yet found the time to do proper commenting, but here's a 
first thought. More details on each UC to follow by the end of next week...

Hannes Pirker schrieb:
> The interesting part of the whole discussion seems to be the point on
> how "broad" or "narrow" the annotation scheme should be.

It seems to me that one critical question is emerging here: the limit of 
what should be and what should not be part of the emotion language.

Hannes pointed out a number of things that go beyond the core task of 
describing the emotion:

> In UC1 there are e.g. attempts to provide labels for
> describing the communicative setting: the individual, the
> interactional situation, the target of the emotion etc.
> In UC2: labels for describing the technical environment: sensors, the
> application, etc.
> In UC3: there are labels for Input Events and Output Events. 

All these pieces of information are certainly important and relevant in 
their respective domain; but only there. In other words, the viability 
of a generic emotion language will depend on a mechanism for including 
(or linking to) domain-specific information where needed, without 
overloading the spec with domain-specific aspects.

How to solve those things technically in XML is a question we can still 
address later (in brief, I see at least three different possible 
solutions: (a) embed emotion-language into a domain-specific language; 
(b) link to external info; or (c) allow for domain-specific sub-structures).

But already now, this question of "what should be part of the language" 
is important when discussing requirements. For example, in the 
requirements document [1] for UC2, Christian has proposed a choice of 
three states for each item:

* definitely “in”	
* could be “in”	
* should be “out”

A different way of phrasing this would be to try and identify things as 
"domain-specific" or "generic" (or, "unclear"...)

For example, in the UC2 document, I would personally consider "ID of 
labeller" as a clear case of "domain-specific" information (the domains 
being something like, "emotion annotation", i.e. irrelevant for 
"generation of emotional system behaviour").

"multiple emotions", or "emotion category", on the other hand, look like 
clear cases of "generic" items to me.

"modalities present" is an "unclear" case to me, at least in the first 
instance: it assumes modalities are relevant, i.e. that we are dealing 
with emotion by a human. Of course this will often be the case, but 
sometimes (e.g., in the "affective reasoner" part of UC3) we are dealing 
with "system-internal" emotions which are not externalised at all; in 
other cases, emotion may be expressed through media that are not human 
modalities (music, colour, font shapes...).

What do you think -- would it make sense to try and classify the various 
proposed items as "domain-specific", "generic", or "unclear"? I think it 
would help us find first parts that we clearly agree upon, and parts 
that need further discussion.


Dr. Marc Schröder, Senior Researcher
DFKI GmbH, Stuhlsatzenhausweg 3, D-66123 Saarbrücken, Germany
Here. Now. Real, first-person experience. Am I there to witness it?
Received on Thursday, 30 November 2006 16:22:14 UTC

This archive was generated by hypermail 2.3.1 : Tuesday, 6 January 2015 19:52:14 UTC