Section: New Results

Lexical Disambiguation

Guy Perrier adapted the methods of lexical disambiguation presented in Mathieu Morey's PhD thesis [49] to the formalism of Tree Adjoining Grammar (TAG) in a common work with Claire Gardent, Yannick Parmentier and Sylvain Schmitz [24] .

More precisely, the algorithm of lexical disambiguation for TAG uses the one-to-one relations between substitution nodes and roots of elementary tress in the parsing process and it takes also into account the position of the subsitution nodes with respect to the anchors in elementary trees, to discard lexical selections that do not respect some constraints. These constraints are implemented through a polarization of the elementary trees and for sake of efficiency, the lexical selections are represented in a compact way with automata.

A major default of the methods of lexical disambiguation presented in Mathieu Morey's PhD thesis is that they ignore local contexts. To overcome this default, Guy Perrier proposed an algorithm to foresee the elementary structures of the grammar that can be inserted between two words that will interact in the parsing process [20] . This algorithm applies to lexicalized grammars, in which the elementary structures are trees.