Corpus design for studying the expression of emotion in speech

In this chapter, I address the issue of emotional speech sampling and the use of corpora in research on emotion expression in different disciplines, including linguistics. I propose that emotions are rare and fleeting events that are difficult to capture in a purely spontaneous fashion, especially as they are likely to be constantly manipulated for the purpose of self-regulation or to fit with social constraints. I propose that emotion sampling design and the use of speech corpora should be adapted to the specific goals pursued by a researcher and the nature of the questions asked. For example, is the work oriented toward an understanding of the production mechanisms underlying emotional speech, that is, from the perspective of the speaker, or does it concern the nature of the recognition from the attribution side, that is, from the perspective of the listener? Or are both perspectives to be employed, with the aim of understanding transmission and communication issues? Is the approach oriented toward an examination of nonlinguistic or extralinguistic cues, or are linguistic and paralinguistic features important? To what extent does the symbolic aspect (iconic, indexical, referential) of the speech feature being studied play a role?