Unaccusative/Unergative Distinction in Turkish: A Connectionist Approach
Cengiz Acartürk, Deniz Zeyrek · 2010
This study presents a novel computational approach to the analysis of unaccusative/unergative distinction in Turkish by employing feed-forward artificial neural networks with a backpropagation algorithm. The findings of the study reveal correspondences between semantic notions and syntactic manifestations of unaccusative/unergative distinction in this language, thus presenting a computational analysis of the distinction at the syntax/semantics interface. The approach is applicable to other languages, particularly the ones which lack an explicit diagnostic such as auxiliary selection but has a number of diagnostics instead. 1