Faites connaître cette offre !
Reference : UPR3251-LAUDEV-001
Workplace : ORSAY
Date of publication : Tuesday, June 16, 2020
Scientific Responsible name : Laurence Devillers
Type of Contract : PhD Student contract / Thesis offer
Contract Period : 36 months
Start date of the thesis : 1 October 2020
Proportion of work : Full time
Remuneration : 2 135,00 € gross monthly
Description of the thesis topic
The thesis will focus on a multi-modal deep learning architecture for detecting social emotions during Human-Human dialogues in emergency call centers. Algorithmic development often rests on the assumption that the input emotions are prototypical emotion expression, which is not true in real-life spoken interactions. This thesis will present methods for interpreting the emotional content of non-prototypical utterances (corpus CEMO) collected in emergency call centers in 2005 and will also test new data coming from another emergency call centers in 2020. Based on emotions detection on both channels voice and words from Automatic Speech Recognition (ASR), we could also try to detect the critical misunderstanding situation between the caller and the operator. The caller's emotional stance, which may range from keeping calm to losing control, is crucial for operators' understanding and assessment of emergency calls. Automatic emotion detection could enrich the platform used by agent in emergency call centers. This thesis is financed within the context of the Chaire AI HUMAAINE : HUman MAchine Affective Interaction & Ethics at LIMSI-CNRS.
A multidisciplinary research laboratory in Orsay (headed by Sophie Rosset), LIMSI brings together 100 researchers and teacher-researchers from the Engineering and Information Sciences as well as the Life Sciences and Human and Social Sciences. The scientific fields thus covered are those of language sciences and technologies, human-computer interaction, virtual and augmented reality, artificial intelligence, fluid and transfer mechanics, and energy.
Constraints and risks
Chaire AI HUMAAINE (LIMSI-CNRS) (septembre 2020-24)
We talk about it on Twitter!