Abstract

Joe Pater's target article calls for greater interaction between neuralnetwork research and linguistics. I expand on this call and show how suchinteraction can benefit both fields. Linguists can contribute to research onneural networks for language technologies by clearly delineating the linguisticcapabilities that can be expected of such systems, and by constructingcontrolled experimental paradigms that can determine whether those desideratahave been met. In the other direction, neural networks can benefit thescientific study of language by providing infrastructure for modeling humansentence processing and for evaluating the necessity of particular innateconstraints on language acquisition.

 

Quick Read (beta)