The SAFE Corpus: illustrating extreme emotions in dynamic situations
Résumé
Existing real-life corpora illustrate everyday life contexts in which social emotions frequently occur. The type of emotional manifestations and the degree of intensity of such emotions are determined by politeness habits and cultural behaviours. This paper shows how the challenge of collecting extreme manifestations of emotion has been addressed with the acquisition of a corpus of fiction, the SAFE Corpus. The aimed application is civil safety and surveillance of public places in particular. A task-dependent annotation strategy is developed with both generic and specific descriptors. A description of the emotional content of the SAFE Corpus is provided. The corpus focuses on the illustration of extreme fear-type emotions in rich and varied contexts. Finally, a detection system of fear emotions based on acoustic cues has been developed to carry out an evaluation.