Characterizing the Errors of Data-Driven Dependency Parsing Models

Abstract

We present a comparative error analysis of the two dominant approaches in datadriven dependency parsing: global, exhaustive, graph-based models, and local, greedy, transition-based models. We show that, in spite of similar performance overall, the two models produce different types of errors, in a way that can be explained by theoretical properties of the two models. This analysis leads to new directions for parser development. 1