Emotion Incubator meeting

21 Aug 2008

See also: IRC log



Catherine, Felix, Enrico, Dylan, Marc


Ian, Bill, Bjoern, Paolo





Chair's summary

There was clear agreement among the participants that we would like to continue working as a subgroup of the MMI working group. Dates of next meeting were agreed. Basic approach proposed for dealing with scale values was agreed, but some questions regarding discrete scales are left open.


1. Future of EMOXG

Marc explains the 3 possibilites for the future of EMOXG (stop, continue as its own working group, join the multimodal interaction as a sub-group)

all the participants of today's meeting agree on the 3rd option (join MM interaction as a sub-group)

Marc will ask in the mailing list if there are other opinions on this issue

2. Next meetings

Phone meetings:

Thursday 4 September, 14:00 UTC

Thursday 18 September, 14:00 UTC

Thursday 2 October, 14:00 UTC


<marc> ask the participants who can come to the meeting in Cannes

3. XML discussion continued: scale value requirements

<marc> Original suggestions in email: http://lists.w3.org/Archives/Public/public-xg-emotion/2008Jul/0008.html

Marc:presents the different issues regarding scales specification: numerical and verbal

Enrico:proposes to restrict the verbal values to 5. Problem of subjectivity arises with using higher number of verbal scales

Instead of verbal discrete scales (using words), it would also be possible to use abstract discrete scales, such as:

<marc> unipolar scale:

<marc> +

<marc> ++

<marc> +++

<marc> ++++

<marc> +++++

<cgi-irc> or:

<cgi-irc> 0

<cgi-irc> +

<cgi-irc> ++

<cgi-irc> +++

<cgi-irc> ++++

Some conclusions were made: we tend to prefer 5 point scales.

Open issue to be discussed: use abstract scale or linguistic scale.


<marc> XML discussion continued, looking at concrete suggestions from email:

Core 3:

<marc> <emotion>

<marc> <dimensions set="Arousal-and-Valence">

<marc> <arousal value="very much" confidence="0.9"/>

<marc> <valence value="slightly positive" confidence="0.3"/>

<marc> </dimensions>

<marc> </emotion>

(Core 4 similar to Core 3)

Regarding sub-element of Core 3, 4: need to verify the values (eg appraisal dimensions) from a repository

Core 5:

<marc> <emotion>

<marc> <action-tendencies set="Frijda">

<marc> <approach activation="(unipolar scale)"/>

<marc> <avoidance activation="(unipolar scale)"/>

<marc> <being-with activation="(unipolar scale)"/>

<marc> ...

<marc> </action-tendencies>

<marc> </emotion>

Core 5: for consistency, we should stick to "value" as an attribute name rather than "activation".

<marc> <emotion>

<marc> <intensity value="(unipolar scale)"/>

<marc> </emotion>

<marc> <emotion>

<marc> <intensity value="0.1" confidence="0.8"/>

<marc> <category set="everyday" name="boredom" confidence="0.1"/>

<marc> </emotion>

Core 7: Intensity should be a separate element rather than an attribute

Confidence (Meta 1):use same type of scale. Which elements should allow for a "confidence" attribute is left for a future discussion.

[End of minutes]

Minutes formatted by David Booth's scribe.perl version 1.133 (CVS log)
$Date: 2008/08/27 14:47:57 $