Affect Detection from Virtual Drama

Abstract

We have developed an intelligent agent to engage with users in virtual drama improvisation previously. The intelligent agent was able to perform sentence-level affect detection especially from user inputs with strong emotional indicators. However, we noticed that emotional expressions are diverse and many inputs with weak or no affect indicators also contain emotional indications but were regarded as neutral expressions by the previous processing. In this paper, we employ latent semantic analysis (LSA) to perform topic detection and intended audience identification for such inputs. Then we also discuss how affect is detected for such inputs without strong emotional linguistic features with the consideration of emotions expressed by the most intended audiences and interpersonal relationships between speakers and audiences. Moreover, uncertainty-based active learning is also employed in this research in order to deal with more open-ended and imbalanced affect detection tasks within or beyond the selected scenarios. Overall, this research enables the intelligent agent to derive the underlying semantic structures embedded in emotional expressions and deal with challenging issues in affect detection tasks.