A Conversational Social Robot for Detecting Emotions Through Drawing Journal Activity

Abstract

This study proposes a conversational social robot with a drawing journal activity that can express children's subjective and honest thoughts and emotions through drawing. The robot leads to conversation during the drawing activity and can detect the child's emotion by analyzing the children's utterance explaining the picture they drew. The children's emotions were classified as one of four emotions which are anger, happiness, sadness and neutral through a BERT-based emotion detecting model. The suitability and readiness of the proposed conversational robot with a drawing activity were then validated by the group of experts in the field of child education.