Identification of emotions by prosody and semantics in French spontaneous speech: A pilot study
Francesca Carbone · Language Mind Culture and Society · 2019
When we talk about emotions, we refer to a complex and structured phenomenon that can influence and be influenced by external and internal elements involved in communication such as context, behavior and feelings of other individuals. During the expression of emotions, these internal and external factors can have effects on both verbal and non-verbal cues which can change according to the context and the intentions of the speakers. This complexity of the emotional phenomenon has induced researchers to focus the attention on individual cues (e.g., intonation, verbal content) in isolation from the others. Although it is possible that the transmission of emotional information is entrusted to a single type of cues, this is certainly the least frequent case in the spontaneous production of emotional speech, in which, usually, the emotional phonological information co-occurs with the other segmental features (e.g. syllables, words). Consequently, studies on individual emotional cues have not increased our understanding of how people integrate nonverbal and verbal cues in the expression of emotions. The aim of this study is to face this topic by assessing the contribution of prosodic parameters (i.e., fundamental frequency, intensity, speech rate) versus verbal content to the decoding of emotions by a perception test in which spontaneous speech is manipulated through the application of filters, validated in previous research. French listeners are asked to identify sentences that were annotated and selected from a French spontaneous corpus (EmoTV). Stimuli are presented in three versions: integral (non-manipulated), verbal (the prosodic information is hidden by applying filters) and nonverbal (the verbal content is filtered and masked). Results display a higher accuracy for integral stimuli suggesting that emotions are decoded by the interaction of the two different channels. This confirms the adequacy of a cognitive model in which verbal and nonverbal cues to emotions are theoretically integrated.