thecenter

Activities and Seminars

Sarah Laszlo. New Language of the Machine: How understanding the brain can help to build better computational models of reading.
 
Date: May 15, 2014

What: New Language of the Machine: How understanding the brain can help to build better computational models of reading.

Where: BCBL auditorium

Who: Sarah Laszlo, Assistant Professor, Binghamton University, State University of New York, Psychology, Linguistics, New York, USA.

When: 12 noon


Parallel Distributed Processing (PDP) models became popular in the 1980s at least in part due to their reliance on neural-style computation. In the subsequent 30 years, PDP models have become extremely influential in understanding reading comprehension, because they allow for interactive processing across multiple levels of representation, as is needed for reading, which relies on (at least) orthographic, phonological, semantic, and syntactic information. While the sophistication of PDP models and the number of behavioral effects they can explain has improved dramatically in the 30 years since their inception, very little work has been done on improving the neural metaphor that in part made them popular to begin with. Here, we discuss several methods for improving the biological plausibility of a classic PDP reading model. We demonstrate that improved neural realism is necessary for simulation of data directly reflecting cortical excitatory and inhibitory post synaptic potentials during reading, as measured using the Event-Related Potential (ERP) technique.