Emin Çelik
Postdoctoral Researcher Max Planck Institute for Software Systems
Date/Time: Tuesday, 09 October 2025, 15:30 pm
Place: SC106 National Magnetic Resonance Research Center Seminar Room
Abstract:
Can language models help us make sense of how our brains process language when we read or listen? In this talk, I will tell you about how we tested this by measuring how well both text- and speech-based language models predict naturalistic fMRI responses. Our results show that text-based language models, unlike their speech-based counterparts, predict brain responses evoked by semantic processing in higher-level language regions. I will then talk about whether large language models (LLMs) can capture word meanings the way humans do. I will show that LLMs generate human-like ratings across a wide range of semantic properties, suggesting they can potentially be used as artificial semantic annotators. I will also discuss how we are extending this approach to annotate a full audiobook and link it to fMRI and MEG data, providing new opportunities to study semantic representation in the brain at scale.
Bio:
Emin Çelik is a postdoctoral researcher at the Max Planck Institute for Software Systems (MPI-SWS) in Germany, where he works on elucidating the principles of semantic representation in the human brain during language comprehension, using tools from natural language processing (including large language models), computational neuroscience, and machine learning. He received his B.S. degree in Electrical and Electronics Engineering from Bilkent University and his M.S. degree in Electrical and Computer Engineering from the University of Texas at Austin. He then returned to Bilkent and joined the newly-established Neuroscience Program where he completed his Ph.D. in 2022 before starting his current position at MPI-SWS