This paper explores the relation between agreement, data quality and machine learning, using the AMI corpus. The paper describes a novel approach that uses contextual information from other modalities to determine a more reliable subset of data, for annotations that have a low overall agreement.
|Name||Lecture Notes in Computer Science|
|Conference||Multimodal Corpora -- From Models of Natural Interaction to Systems and Applications|
|Period||27/05/08 → 27/05/08|
- EC Grant Agreement nr.: FP6/033812