Multiword Units in an MT Lexicon
Tamás Váradi · Repository of the Academy's Library (Library of the Hungarian Academy of Sciences) · 2006
Multiword units significantly contribute to the robustness of MT systems as they reduce the inevitable ambiguity inherent in word to word matching.The paper focuses on a relatively little studied kind of MW units which are partially fixed and partially productive.In fact, MW units will be shown to form a continuum between completely frozen expression where the lexical elements are specified at the level of particular word forms and those which are produced by syntactic rules defined in terms of general part of speech categories.The paper will argue for the use of local grammars proposed by Maurice Gross to capture the productive regularity of MW units and will illustrate a uniform implementation of them in the NooJ grammar development framework.