A Computational Theory for the Emergence of Grammatical Categories in Cortical Dynamics

THIRUVATHUKAL, GEORGE; RIZZI, SILVIO; WAINSELBOIM, ALEJANDRO; DEMATTÍES, DARÍO; PÉREZ, MAURICIO DAVID

Buenos Aires

Congreso; XXXV Annual Meeting of the Argentinian Society for Neuroscience Research; 2020

Sociedad Argentina de Investigación en Neurociencias

A general agreement in psycholinguistics claims that syntax and meaning are unifiedprecisely and very quickly during online sentence processing.Although several theories have advanced arguments regarding the neurocomputationalbases of this phenomenon, we argue that these theories could potentially benefit byincluding neurophysiological data concerning cortical dynamics.In this work we introduce a computational model inspired in the dynamics of corticaltissue.In our model, proximal afferent dendrites produce stochastic cellular activations, whiledistal dendritic branches contribute independently to somatic depolarization by means ofdendritic spikes, and finally, prediction failures produce massive firing events preventingformation of sparse distributed representations.This model combines semantic and syntactic constraints for each word in a sentencecontext until grammatically related word function discrimination emerges spontaneouslyby the sole correlation of lexical information from different sources without applyingcomplex optimization methods.We show that the sparse activation features returned by our approach are well suited toaccomplish grammatical function classification of individual words in a sentence. In thisway we develop a biologically guided computational explanation for linguisticallyrelevant unification processes in cortex which connects psycholinguistics toneurobiological accounts of language.