Annotators of multimodal corpora rely on a combination of audio and video features to assign labels to the events observed. The reliability of annotations may be influenced by the presences or absence of certain key features. For practical applications it can be useful to know what circumstances determined fluctuations in the interannotator agreement. In this paper we consider the case of annotations of addressing on the AMI corpus.
|Title of host publication||Proceedings of the LREC Workshop on Multimodal Corpora|
|Editors||J-C. Martin, P. Paggio, M. Kipp, Dirk K.J. Heylen|
|Place of Publication||Paris, France|
|Number of pages||4|
|Publication status||Published - May 2008|
- HMI-MI: MULTIMODAL INTERACTIONS
- EC Grant Agreement nr.: FP6/033812