Re: [EMOXG] Deliverable report published as first draft: Emotion Markup Language: Requirements with Priorities

Kostas, Catherine, all,

Kostas Karpouzis schrieb:
> 
> Catherine Pelachaud wrote:
>>
>>
>> My 2 cents on the definition of min, max.
>> In MPEG-4 facial Animation Parameters have no min-max. Any values are 
>> allowed. The difficulty is to make sure that all MPEG-4 player 
>> interprets the values in a similar manner. To ensure this, detailled 
>> examples are provided as well as animation files that served as test bed.
>> I also like the idea of not having min-max specified. It allows for 
>> much more flexibility and also not to have to define what it is 
>> absolute max 
> Plus, some applications may _want_ to use excessive values (e.g. for 
> eye-popping, cartoon-like animation) In addition to this, in MPEG-4 
> units are inherent in the measurement, since values normalized wrt 
> constant distances; e.g. FAPs related to the eyebrows are normalized 
> using the distance between the eyes and the nose, which (normally) is a 
> constant.

Indeed, we need some sort of reference so that values can be 
interpreted. If I understand correctly, the reference is built into the 
MPEG model as a normalised length defining facial geometry. We could do 
the same, stating that, e.g., 1 (or 100, I don't care) is the maximum 
intensity that is normally expected. If someone wants to exaggerate, 
then we can allow for values higher than 1 -- in such cases, then, it 
should be clear that "unnatural" emotional properties are being modelled.

> Regarding fuzziness and labels, I would think that it's best to leave it 
> up to applications reading and using EMOXG to define them or, in any 
> case, a higher-level structure which again would be app-dependent (an 
> ontology relating feature points to FAPs to expressivity maybe?) In some 
> cases, or for certain users, a specific measurement of eyebrow movement, 
> for instance, may correspond to 'high' activation, while in other cases 
> or contexts, the same measurement may be labeled 'medium'

This example seems to me outside of the EmotionML: you describe the 
question of how to interpret a given expressive behaviour in terms of 
emotion.

If we go for qualitative scale values, then I would suggest very much 
that we try to agree on a set of labels if possible. Only if there is a 
strong need, we should make the set of labels itself flexibly 
specifiable. Anyway, this is for the spec discussion, not the 
requirements doc.

Best,
Marc

-- 
Dr. Marc Schröder, Senior Researcher at DFKI GmbH
Coordinator EU FP7 Project SEMAINE http://www.semaine-project.eu
Chair W3C Emotion ML Incubator http://www.w3.org/2005/Incubator/emotion
Portal Editor http://emotion-research.net
Team Leader DFKI Speech Group http://mary.dfki.de
Project Leader DFG project PAVOQUE http://mary.dfki.de/pavoque

Homepage: http://www.dfki.de/~schroed
Email: schroed@dfki.de
Phone: +49-681-302-5303
Postal address: DFKI GmbH, Campus D3_2, Stuhlsatzenhausweg 3, D-66123 
Saarbrücken, Germany
--
Official DFKI coordinates:
Deutsches Forschungszentrum fuer Kuenstliche Intelligenz GmbH
Trippstadter Strasse 122, D-67663 Kaiserslautern, Germany
Geschaeftsfuehrung:
Prof. Dr. Dr. h.c. mult. Wolfgang Wahlster (Vorsitzender)
Dr. Walter Olthoff
Vorsitzender des Aufsichtsrats: Prof. Dr. h.c. Hans A. Aukes
Amtsgericht Kaiserslautern, HRB 2313

Received on Tuesday, 13 May 2008 07:14:20 UTC