Exploring Sounded and Silent Laughter in Multiparty Task-based and Social Interaction - Audio, Video and Biometric Signals

We report on our explorations of laughter in multiparty spoken interaction. Laughter is universally observed in human interaction. It is multimodal in nature: a stereotyped exhalation from the mouth in conjunction with rhythmic head and body movement. Predominantly occurring in company rather than solo, it is believed to aid social bonding. Spoken interaction is widely studied through corpus analysis, often concentrating on ‘taskbased’ interactions such as information gap activities and real or staged business meetings. Task-based interactions rely heavily on verbal information exchange while the immediate task in natural conversation is social bonding. We investigate laughter in situ in task-based and social interaction, using corpora of non-scripted (spontaneous) multiparty interaction: the task-oriented AMI meetings corpus, and the conversational TableTalk, and D-ANS corpora. We outline extension of previous work on laughter and topic change, describe the collection of natural social spoken interaction in the DANS corpus including audio-visual and biosignals, and describe an annotation experiment on multimodal aspects of laughter. We discuss the results and signal current and future research directions.