Speech Science Forum -- Andrew Thwaites
15 May 2025, 3:00 pm

Inferring the pathways of spoken language comprehension through Audio-LLM alignment
Event Information
Open to
- All
Availability
- Yes
Organiser
-
Victor Rosi
Location
-
G152Wakefield StreetLondonWC1N 1PF
Language comprehension is a complex cognitive process, and modelling it accurately has long been a challenge. The advent of modern Artificial Neural Networks (ANNs), capable of transcribing speech and other stimuli with human-level accuracy, provides researchers with the opportunity to compare the these models with neural processing, a topic known as ANN-brain alignment.
This talk will explore methods for assessing the alignment between ANNs trained for spoken language comprehension and neural activity, with a focus on magnetoencephalography (MEG) and electroencephalography (EEG) data. It will cover both neuroimaging methods and insights into artificial and biological language processing mechanisms.
zoom link : https://ucl.zoom.us/j/92052680901
About the Speaker
Andrew Thwaites
Principal Research Fellow at University College London
Andrew Thwaites is a researcher in magnetoencephalography (“MEG”) and electroencephalography (“EEG”), based at SHaPS, in the Division of Psychology and Language Sciences, UCL. He focuses primarily on methods related to the mapping of audition, vision, and language in the human brain.
More about Andrew Thwaites