Emotion-Oriented Systems

Chapter 3

Emotional Corpora: from Acquisition to Modeling ¹

3.1. Introduction

The purpose of affective systems is to incorporate emotions into computational systems. Central to this is the creation of emotional corpora. They are essential both for creating human–machine interaction systems that can manage and regulate emotional responses and for carrying out perceptive studies into emotion. Corpora are a body of consistent and significant recorded data that we can use to create and design models. The composition and interpretation of corpora obviously depends on the objective for modeling. For example, generating emotional expressions for a conversational agent could require an extensive corpus of emotional behavior from a single locutor chosen for their emotional expressiveness. In contrast, an application designed to detect emotion needs to be highly sensitive to different interlocutors and therefore requires a large multi-speaker corpus.

Emotions are a complex dynamic process [COP 09] and manifest themselves differently from one person to another. The brain is the source of multiple emotions that evolve dynamically mixing pleasant and unpleasant, conscious and unconscious emotions that are stimulated by external or internal events. An emotion is a state of consciousness that can be pleasant or upsetting, which is accompanied by a bodily reaction when faced with an external or internal event (e.g. noise, light, memory or concept). For example, anger makes blood rush to the ...