Anvil implementation of the MUMIN 4.2 coding scheme, created for the Nordic Multimodal Corpus Analysis in the Nordic Countries annotations, July 2009. Based on the MUMIN 3.3. coding scheme created by Jens Allwood, Loredana Cerrato, Laila Dybkær, Kristiina Jokinen, Costanza Navarretta and Patrizia Paggio. See original coding scheme.
New in this version: a Body-Interlocutor-type has been introduced, combined semiotic types have been removed, a few errors in the comments have been corrected, and the speaker tracks have been simplified only to include the following features:
plus a number of multimodal relations of type multilink.
The other features from the MUMIN scheme are not deleted from the tracks but only commented out.
The purpose of this version of the MUMIN coding scheme is to allow functional annotation of multimodal expressions, as well as the way in which expressions belonging to different modalities are combined. The functions defined are:
Feedback, turn management and sequencing are formalised similarly to what done in MUMIN 3.3. However, the current feature taxonomy is not totally equivalent. In the old scheme, understanding and acceptance were subtypes of FeedbackGive and FeedbackElicit. Now, the scheme has a basic FeedBack feature (CPU/Other) in combination with a FeedbackDirection, where the distinction between FeedbackGive and Elicit is captured, as well as an Agreement feature.
However, only feedback features are included in the speaker tracks to simplify the coding procedure.
Tracks are specified to code speech and gestures of two dialogue participants, here called Speaker A and Speaker B. The names can of course be changed, and more tracks may be added. In this version of the coding scheme, the tracks only include a subset of the MUMIN features for facial displays, and features for body posture.
| Nodes | |
wave
Container
|
--- |
pitch
Container
|
--- |
SpeakerA
Group
|
--- |
SpeakerB
Group
|
--- |