An Interface to Simplify Annotation of Emotional Behaviour

Research in affective computing is increasingly moving towards naturalistic data. Capturing and annotating such complex data is a massively challenging task. This paper describes a simple and efficient annotation scheme that promotes context-sensitive data labelling via an easy to use interface in an attempt to reduce reliance on expert or trained coders. Additionally, the same labelling interface can be used to obtain self-report of emotional behaviour from subjects. This annotation method has been designed to allow faster labelling of data with a minimal learning curve as part of our research in studying non-verbal expressivity of affect in computer based learning environments and is currently being evaluated. Design decisions are based on feedback from usage of initial prototype as well as relevance to our domain of interest. We anticipate that this can enable faster preparation of representative data in an effective manner for use in automatic analysis studies.