Jixing Li & John Hale
Lecture 2 | Leipzig Lectures on Language—Combinatorics 2021
Neuro-computational models of language processing: The case of reference and coreference
Efforts to map between the abstract constructs of linguistics and concrete properties of the human brain face significant challenges. In this talk, we first outline one approach that relies on computational models of language processing to answering the Mapping Problem. We then present one empirical study that applies the neuro-computational approach to understand the brain mechanisms of referential processing. Using both fMRI and MEG, we identified a consistent increase of activity in the left middle temporal gyrus and the angular gyrus for pronoun processing during naturalistic listening for both English and Chinese speakers. We then evaluated the neural fit of three symbolic models that each formalizes a different strand of explanation for pronoun resolution in the cognitive and linguistic literature, as well as two deep neural network models with an LSTM or a Transformer architecture. Our results favor the memory-based ACT-R model, suggesting a domain-general mechanism of pronoun resolution that resembles memory retrieval.
About the speakers
Jixing Li is postdoctoral researcher in the Neuroscience of Language Lab (NeLLab) at NYU Abu Dhabi. Her research applies computational models to understand how the human brain represents and computes semantic and syntactic information during language comprehension.
John Hale is Professor at the Linguistics Department, University of Georgia. His work centres on language comprehension using analysis tools from computational linguistics and corpus methods.
Keywords: language processing; incremental comprehension; phrase structure parsing; semantic composition; computational linguistics