Study of Social-Affective Communication: Emotion Recognition, Emotional Triggers Prediction, and Dialogue Response Selection

Nurul Fithria Lubis (1551128)


In a social setting, emotion works in two ways: a person is expressing their emotion, and also affected by their conversational counterpart. This creates a rich and dynamic social interaction between humans. It is argued that humans also impose emotional aspect when interacting with computers and machines. However, the majority of existing works have not yet fully considered the two-way role of emotion: the influence of others on how a person's emotion changes and fluctuates in an interaction. In this work, we attempt to utilize such knowledge to elicit positive emotion in Human-Computer Interaction (HCI). We identify the processes that amounts to the social-affective loop: emotion expression, recognition, emotional triggers, and response. Accordingly, we approach our objective incrementally through three main tasks: 1) Recognizing affective states, 2) predicting social-affective events, and 3) eliciting a positive emotional response. Furthermore, as emotional corpus is pre-requisite in each of these tasks, we construct a corpus of spontaneous social-affective interaction between humans. We successfully construct the components to perform each of the mentioned tasks. These components will serve as the building blocks for future integration into a system that will be able to offer emotional support through real-time interaction with users.