Emmanuel Dupoux
استاد · Language Acquisition
Swiss Federal Institute of Technology in Lausanneمعرفی
Emmanuel Dupoux is a Professor at École des Hautes Études en Sciences Sociales (EHESS), affiliated with the School of Advanced Studies in Social Sciences and the Laboratory of Cognitive Science and Psycholinguistics (LSCP). His work bridges cognitive science, computational linguistics, and machine learning to study infant language acquisition and social cognition.
- Co-creator and director of EHESS Cognitive Science Master program
- Former LSCP laboratory director (1998-2009)
- Current research focuses on textless speech modeling and ecological audio analysis
Research areas include:
- Modeling early language acquisition mechanisms using Bayesian models and HMM
- Investigating phonological 'deafness' and critical age plasticity
- Studying social cognition development through infant-toddler experiments
His technical contributions span speech processing toolkits like Shennong, emergent communication frameworks (STOP dataset), and innovative approaches to self-supervised speech modeling. Recent publications demonstrate breakthroughs in:
- Topography-inspired CNN designs for better memory efficiency
- Prosody-aware generative spoken language models (pGSLM)
- Textless emotion conversion systems
He actively explores how machine learning can reverse-engineer infant language learning processes from ecological audio data, with applications in neurodegenerative disease diagnostics.
Key collaborations include:
- INRIA's Cognitive Machine Learning (CoML) team
- Facebook AI Research (FAIR) partnerships
- VoxPopuli multilingual speech corpus development
