Multiword Expression Identification with Tree Substitution Grammars: A Parsing tour de force with French

Spence Green (1) , Marie-Catherine de Marneffe (2) , John Bauer (1) , Christopher D. Manning (1, 2)

1 Computer Science Department [Stanford]
2 Linguistics Department

Multiword expressions (MWE), a known nui-sance for both linguistics and NLP, blur the lines between syntax and semantics. Previous work on MWE identification has relied primar-ily on surface statistics, which perform poorly for longer MWEs and cannot model discontin-uous expressions. To address these problems, we show that even the simplest parsing mod-els can effectively identify MWEs of arbitrary length, and that Tree Substitution Grammars achieve the best results. Our experiments show a 36.4% F1 absolute improvement for French over an n-gram surface statistics baseline, cur-rently the predominant method for MWE iden-tification. Our models are useful for several NLP tasks in which MWE pre-grouping has improved accuracy.

Mots clés: multiword expression; lexicon-grammar

Domaines: Traitement du texte et du document; Linguistique

Spence Green, Marie-Catherine de Marneffe, John Bauer, Christopher D. Manning. Multiword Expression Identification with Tree Substitution Grammars: A Parsing tour de force with French. Conference on Empirical Methods in Natural Language Processing, Jul 2011, Edinburgh, United Kingdom. ⟨hal-01111383⟩