Finite Matters

  • Authors:
  • Lilja Øvrelid

  • Affiliations:
  • NLP-unit, Dept. of Swedish, University of Gothenburg,

  • Venue:
  • GoTAL '08 Proceedings of the 6th international conference on Advances in Natural Language Processing
  • Year:
  • 2008

Quantified Score

Hi-index 0.00

Visualization

Abstract

This paper investigates the effect of a set of verbal features in data-driven dependency parsing of Swedish. Following an error analysis of a baseline parser, we show that the addition of information on verbal features such as tense and voice can give significant improvements over this baseline and, in particular, in the analysis of syntactic arguments. We furthermore show the importance of the binary property of finiteness for the parsing of Scandinavian and demonstrate that highly similar effects may be achieved with automatically acquired information.