Deep learning algorithms, natural language processing, and the brain, with Jean-Rémi King
The Language Neuroscience Podcast
00:00
Long-term Predictions and Hierarchical Structures in Language Processing
The chapter explores the difference between large language models and humans in processing language, focusing on the need for nonlinear composition in understanding the meaning of phrases and sentences. It discusses the importance of training language models to predict multiple levels of representation and introduces a forecast window into the model to improve the correspondence with the brain's activations.
Transcript
Play full episode
