On computational simulations of acquisition of natural language syntax

Andrzej Nowak , Piotr Kochański , Peter W. Culicover , Wojciech Borkowski

Abstract

In this article we present various distributional approaches to language acquisition. These are based on statistical parsing of sentence corpora and model certain aspects of the process of language learning. Our aim is to give a non-specialist in this field a general idea of when such methods are useful, and what their advantages and limitations are. We show how distributional models have been applied to fundamental linguistic questions about the innate properties of human language. We also discuss possible directions of development of distributional methods, so they might be able to account for difficult linguistic processes such as the acquisition of syntax.