9th Annual Conference of the International Speech Communication Association

Brisbane, Australia
September 22-26, 2008

Scripted Dialogs versus Improvisation: Lessons Learned About Emotional Elicitation Techniques from the IEMOCAP Database

Carlos Busso, Shrikanth S. Narayanan

University of Southern California, USA

Recording high quality data is an important step in the study of emotions. Given the inherent limitations and complexities of the current approaches to capture natural emotions collected in real-life scenarios, the use of professional actors appears to be a viable research methodology. To avoid stereotypical realization of the emotions, better elicitation techniques rooted in theatrical performance are needed. Based on the lessons learned from the collection of the IEMOCAP database, this paper analyzes the advantages and disadvantages of two of the most appealing elicitation approaches: improvisation, and scripted dialogs. These methods are studied in terms of the lexical content, disfluencies, facial activity and emotional content. The results indicate that spontaneous sessions have higher levels of disfluencies and overlapped speech. Also, the emotional content seems to be more intense than in scripted sessions, as revealed in subjective evaluations.

Full Paper

Bibliographic reference.  Busso, Carlos / Narayanan, Shrikanth S. (2008): "Scripted dialogs versus improvisation: lessons learned about emotional elicitation techniques from the IEMOCAP database", In INTERSPEECH-2008, 1670-1673.