This paper presents the development of the Emotive resource, a corpus of spoken French designed for the study of emotional speech, and its initial exploration. The study presented here is focused on affective syntax. After setting out our hypotheses regarding the syntactic marking of emotional speech, we describe the methodology used for collecting the data and metadata, the selection of sequences to be annotated for the study of syntactic markers, and the segmentation issues raised by delimiting emotional sequences within the corpus. We then justify the descriptors chosen for annotating these sequences. Finally, we present our statistical results, which reveal certain significant differences between emotional speech and so-called “neutral” speech on the one hand, and between sequences with positive valence and those with negative valence on the other. These results allow us to confirm the role of syntax in emotional communication, while also distinguishing, among the emotional patterns identified, the significant contrasts between sequences with negative valence and those with positive valence.
Hirst et al. (Mon,) studied this question.