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


From: Hannes Pirker <hannes@ofai.at>
Date: Thu, 26 Oct 2006 21:03:42 +0200 (CEST)
Message-Id: <200610261903.k9QJ3gl3003195@rose.ofai.at>
To: public-xg-emotion@w3c.org

Dear [EMOXG] colleagues 

Here is the use case that is related to the work at OFAI.

I am sending this out as email, but it is as well added to the
UseCase-document at our WIKI.

This UC obviously relates to TUC1. I added it as new item 1h right now. but
suggest it could be merged with to 1c (JeanClaude) as the requirements
in my opinion are a sub-set of those issued by JeanClaude.

Tanja recorded a video corpus where actors under the supervision of a
director were instructed to produce isolated sentences with 10
different (categorically defined) emotions. In addition some of these
emotions had to be produced with i) increased intensity ii) decreased
intensity and iii) as the person would try to (unsuccessfully)
hide/suppress her emotion. 

This way for each sentence its *intended* emotion, intensity and
possible regulation attempts are already known. In a next step ratings
of human annotators are added, who are rating the quality of the
actors' performance: i) on how well the intended emotional content can
be actually perceived (i.e. this is some skewed variant of 'annotator
confidence') and ii) a rating of on how believability and naturalness
of the performance. 

In the future extracts of the corpus should be used in classical
rating experiments. These experiments may be performed on different
combinations of modalities (i.e. full-body video, facial video, each
with and without speech).

 a video clip (typically containing one sentence)

emotion description:
 (intended) emotion categories
 (intended) intensity of an emotion 
 (intended) regulation of an emotion (only 'suppress' is used)
 labeller rating on the degree of perceivability of the intended emotion
 labeller rating on the degree of naturalness of the performance.

  information on the modality (face, voice, body) used as basis for ratings

Hannes Pirker  -- Austrian Research Inst. for Artificial Intelligence     --
hannes(DOT)pirker~AT~ofai.at +43/1/532 4621-3  www.ofai.at/~hannes.pirker --
Received on Thursday, 26 October 2006 22:33:37 UTC

This archive was generated by hypermail 2.3.1 : Tuesday, 6 January 2015 19:52:13 UTC