Corpus-based Grammar Specialization
Nicola Cancedda, Christer Samuelsson
Broad-coverage grammars tend to be very ambiguous. When such a grammar has to be used on a restricted
domain, it can be desirable to specialize it, i.e. to trade part of the coverage for a reduction in ambiguity.
This paper presents a novel formulation of grammar specialization as an optimization problem, in which the
search is guided by a global measure combining coverage and ambiguity. The method, applicable to any
unification grammar with a phrase-structure backbone, is shown to be effective in specializing a
broad-coverage LFG for French.
Proc. of the Fourth Conference on Computational Natural Language Learning (CoNLL'2000), Lisbon, Portugal, 2000
conll00.pdf (182.59 kB)
conll00.ps (204.35 kB)