Towards a dependency parser for Greek using a small training data set

  1. Jesús Herrera de la Cruz
  2. Pablo Gervás Gómez-Navarro
Journal:
Procesamiento del lenguaje natural

ISSN: 1135-5948

Year of publication: 2008

Issue: 41

Pages: 29-39

Type: Article

More publications in: Procesamiento del lenguaje natural

Abstract

Some experiments have been accomplished in order to determine strategies that should be followed to build a small corpus capable to train accurately a dependency parser for Greek, using a Machine Learning tool. Thus, several problems that should be treated such syntactic coverage, effect of word order or effect of morphology in languages with syntactic roles expressed morphologically, are empirically studied. With the results presented we would like to lay the foundations for a systematic and effective way to develop dependency parsers when lacking huge training corpora. The ideas outlined could be used not only for Greek but for other languages.