Advanced search
Start date
Betweenand

Computing language with mirror neurons

Grant number: 16/18825-4
Support type:Research Grants - Research Partnership for Technological Innovation - PITE
Duration: August 01, 2017 - July 31, 2019
Field of knowledge:Biological Sciences - Biophysics
Cooperation agreement: IBM Brasil
Principal Investigator:Fábio Marques Simões de Souza
Grantee:Fábio Marques Simões de Souza
Home Institution: Centro de Matemática, Computação e Cognição (CMCC). Universidade Federal do ABC (UFABC). Ministério da Educação (Brasil). Santo André, SP, Brazil
Company: IBM Brasil - Indústria, Máquinas e Serviços Ltda
Co-Principal Investigators:Maria Teresa Carthery Goulart
Associated grant(s):17/50406-4 - Computing language and cognitive deficits in a three layered cortex, AP.R

Abstract

Computational modeling of the neural correlates of language processing is a fundamental step toward the cognitive computing of natural language. However, most studies of the neural correlates of text reading focus on nouns, and little is known about the underlying neural mechanisms for processing and recognizing verbs. Thus, our aim is to develop a neural network model with two layers that recognizes verbs accordingly with their lexical parameters. We hypothesize that neurons from Broca's and Wernicke's areas of language work similarly to mirror neurons that fire action potentials both when the actor executes an action or when the actor observes someone else executing the same action. In this way, Broca's area mirror neurons would fire action potentials either during production of motor actions -speaking, signing or writing- that encode a verb, or when Wernicke´s area pre-synaptic neurons are activated by parameters that identify that verb. The input and output neurons will be connected by synapses with spike-timing dependent plasticity to yield mirroring function. Thus, output neurons should fire either to initiate a specific action that they encode or when observing a word that represents that action. The successful network should be able to mimic the cognitive performance of human subjects on a task of verb recognition. (AU)

Distribution map of accesses to this page
Click here to view the access summary to this page.