Affective State Generation, Transition and Facial Expression for a Social Robot Using Texts and Personality Traits

Abstract

Studies focusing on generating and expressing appropriate emotions in robots based on their utterances play a crucial role in achieving natural human-robot social interaction. Existing research predominantly emphasizes generating reactive emotions in robots that correspond to users’ emotions, typically identified through visual cues like facial expressions. However, these approaches may have limitations in effectively conveying robot emotions in applications such as book reading services and robot dialogues, which do not necessarily rely on the user’s emotional input. To address these gaps, we propose a novel methodology for generating emotions and expressions in robots that align with their utterances (i.e., texts), while considering the emotional transitions observed in humans. Our method represents emotions within the affective state space of pleasure and arousal. Initially, sentiment analysis is applied to estimate the affective state of a robot text within this state space. A transition rule is defined within the same space. The robot’s emotion and face expression are updated in accordance with this transition rule, taking into account the affective state derived from sentiment analysis and the pre-defined personality of the robot as specified by the user. To evaluate our approach, we conducted a scenario-based experiment. The results demonstrated that the robot’s emotions and corresponding face expression could change along the path determined by the transition rule. Moreover, different paths were generated depending on the specific personality traits assigned to the robot. By incorporating these elements, the proposed method could enhance the naturalness and effectiveness of human-robot interaction.