Daniela Perekopska - Catherine Mathon

EA 333, ARP, Université Paris 7

8. spoken communication, emotions, feelings and attitudes

Played versus spontaneous emotion: a perceptible difference ?

Perekopska Daniela, Almost all the studies about emotions, including the most famous, are based on a non-natural corpus, played by actors, (or an elicited speech corpus). The sentences are often chosen in order to be semantically neutral, built with written syntax. Nonetheless, more and more researchers are interested in spontaneous emotion, and are trying to obtain interesting corpus under natural conditions. The purpose/aim of our study is to show that there is a perceptive difference between spontaneous emotion and played emotion. Our hypothesis is that differences between played emotion and spontaneous emotion can be perceived in speech, because of prosody as well as syntax. Indeed, oral speech has got its own syntax, very different of written syntax, especially because of disfluencies that mark the spontaneous discourse. Our corpus is composed of : • natural sentences extracted from a corpus of spontaneous speech that contains anger, • the same sentences where disfluencies (sound pauses, repetitions, hesitations, false departures) were removed. What is more, we asked 3 French speakers to play sentences : • with exactly the same content as the original ones, • then with the same content without disfluencies, • finally, sentences reworked with written syntax. We wanted to test how oral marks could influence perception of emotion as natural or not. We built a perceptual test with these five types of stimuli. We tested this corpus on 10 French subjects, in order to confirm our hypothesis, and to find which parameter is the most important, prosody or syntactic organization, to differentiate between played anger and real anger. Subjects just had to listen to the stimuli and to say if the sentence they had listened to was played or spontaneous. In addition to the perceptual test, we made a contrastive acoustical analysis (F0, intensity, speech rate) between the natural stimuli extracted from spontaneous corpus, and the sentences produced by our French speakers, in order to observe the prosodic differences.