Grammar Error Detection with Best Approximated Parse

Jean-Philippe Prost (1)

1 LIFO - Laboratoire d'Informatique Fondamentale d'Orléans

In this paper, we propose that grammar error detection be disambiguated in generating the connected parse(s) of optimal merit for the full input utterance, in overcoming the cheapest error. The detected error( s) are described as violated grammatical constraints in a framework for Model- Theoretic Syntax (MTS). We present a parsing algorithm for MTS, which only relies on a grammar of well-formedness, in that the process does not require any extragrammatical resources, additional rules for constraint relaxation or error handling, or any recovery process.

Domaines: Informatique et langage [cs.CL]; Traitement du texte et du document

Jean-Philippe Prost. Grammar Error Detection with Best Approximated Parse. 11th International Conference on Parsing Technology (IWPT'09), Oct 2009, Paris, France. pp.172--175. ⟨hal-00468007⟩