Interpreting Ambiguous Emotional Expressions

Emotion expression is a complex process involving de-pendencies based on time, speaker, context, mood, person-ality, and culture. Emotion classification algorithms de-signed for real-world application must be able to inter-pret the emotional content of an utterance or dialog given the modulations resulting from these and other dependen-cies. Algorithmic development often rests on the assump-tion that the input emotions are uniformly recognized by a pool of evaluators. However, this style of consistent proto-typical emotion expression often does not exist outside of a laboratory environment. This paper presents methods for interpreting the emotional content of non-prototypical ut-terances. These methods include modeling across multiple time-scales and modeling interaction dynamics between in-terlocutors. This paper recommends classifying emotions based on emotional profiles, or soft-labels, of emotion ex-pression rather than relying on just raw acoustic features or categorical hard labels. Emotion expression is both in-teractive and dynamic. Consequently, to accurately recog-nize emotional content, these aspects must be incorporated during algorithmic design to improve classification perfor-mance. 1