Abstract

In this talk, we compare higher order graph-based and transition-based parsing approaches in terms of their properties and give some background on the development in the last decade. The differences in those two approaches have lead to number of combination approaches to gain the best of both worlds. We briefly compare these advanced parsers with constituent parsers that perform often competitive in terms of accuracy despite more than a decade of intensive research into dependency parsing.

Recently, dependency-parsing systems aim to join morphological and syntactic analysis. Starting from a transition-based model for joint part-of-speech tagging and dependency parsing, we explore different ways of integrating morphological features into the model. We also investigate the use of rule-based morphological analysers and the use of word clusters to tackle the sparsity of lexical features. We will present an evaluation on five morphologically rich languages that shows consistent improvements in both morphological and syntactic accuracy for joint prediction over a pipeline model. The final results improve the state of the art in dependency parsing for these languages.