On the Effect of Head Tagging on Parsing Discontinuous Dependencies in French

Ophélie Lacroix (1)

1 LINA - Laboratoire d'Informatique de Nantes Atlantique

In this paper we aim at showing the strong impact of head tagging on syntactic dependency parsing. The rules of categorial dependency grammar used to parse French deal with discontinuous dependencies and long distance syntactic relations. Such parsing method produces a substantial number of dependency structures and requires too much parsing time. We show that a local tagging method can reduce these problems and help to solve the global problem of dependency parsing disambiguation. Then we adapt a tagging method (CRF) to types of the categorial dependency grammar. We obtain a dependency head preselection allowing us to reduce parsing ambiguity and to see that we can find distant relation of dependencies through local results of such method.

Domains: Machine Learning [cs.LG]; Document and Text Processing

Ophélie Lacroix. On the Effect of Head Tagging on Parsing Discontinuous Dependencies in French. StuS 2013, Aug 2013, Düsseldorf, Germany. pp.93-103. ⟨hal-00918354⟩