Affect Detection from Semantic Interpretation of Drama Improvisation

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 from user inputs with strong emotional indicators. However, we noticed that many inputs with weak or no affect indicators also contain emotional implication but were regarded as neutral expressions by the previous interpretation. In this paper, we employ latent semantic analysis to go beyond linguistic restrictions and to perform topic theme detection and identify target audiences for those inputs with vague affect indicators and ambiguous target audiences. We also discuss how emotions embedded in such emotionally ambiguous inputs are detected with the consideration of interpersonal relationships, special sentence types and emotions experienced by the target audiences using a neural network based contextual affect detection. The work contributes to the conference themes on discourse and pragmatics, semantics and sentiment and text classification. TITLE AND ABSTRACT IN CHINESE