Issues in data labelling

Abstract

Labelling emotion databases is not a purely technical matter. It is bound up with theoretical issues. Different issues affect labelling of emotional content, labelling of the signs that convey emotion, and labelling of the relevant context. Linked to these are representational issues, involving time course, consensus and divergence, and connections between states and events. From that background comes a wealth of resources for labelling emotion, involving not only everyday emotion words but also affect dimensions, and labels for combination types, appraisal categories, and authenticity. Resources for labelling signs of emotion cover linguistic, vocal, face descriptors, plus descriptors for gesture, and relevant physiological variables. Resources for labelling context are developing.