Exploring Different Approaches for Parsing Telugu

Abstract

In this paper we explore different approaches for parsing Telugu. We consider three popular dependency parsers namely, MaltParser, MSTParser and TurboParser. We first experiment with different parser and feature settings and show the impact of different settings. We then explore different ways of ensembling these parsers. We also provide a detailed analysis of the performance of all the approaches on major dependency labels and different distance ranges. We report our results on test data of Telugu dependency treebank provided in the ICON 2010 tools contest on Indian languages dependency parsing. We obtain state-of-the art performance of 91.8% in unlabelled attachment score and 70.0% in labelled attachment score.