Emotion and Its Triggers in Human Spoken Dialogue: Recognition and Analysis

Abstract

Human communication is naturally colored by emotion, triggered by the other speakers involved in the interaction. Therefore, to build a natural spoken dialogue system, it is essential to consider emotional aspects, which should be done not only by identifying user emotion, but also by investigating the reason why the emotion occurred. The ability to do so is especially important in situated dialogue, where the current situation plays a role in the interaction. In this paper, we propose a method of automatic recognition of emotion using support vector machine (SVM) and present further analysis regarding emotion triggers. Experiments were performed on an emotionally colorful dialogue corpus. The result shows performance that surpasses random guessing accuracy.