Extension lexicale de définitions grâce à des corpus annotés en sens

Abstract : Lexical Expansion of definitions based on sense-annotated corpus For many natural language processing tasks and applications, it is necessary to determine the semantic relatedness between senses, words or text segments. In this article, we focus on a knowledge-based measure, the Lesk measure, which is certainly among the most commonly used. The similarity between two senses is computed as the number of overlapping words in the definitions of the senses from a dictionary. In this article, we study the expansion of definitions through the use of sense-annotated corpora. The idea is to take into account words that are most frequently used around a particular sense and to use the top of the frequency distribution to extend the corresponding definition. We show better performances on a Word Sense Disambiguation task surpassing state-of-the-art
Document type :
Conference papers
Liste complète des métadonnées

Cited literature [25 references]  Display  Hide  Download

https://hal.archives-ouvertes.fr/hal-01332850
Contributor : Didier Schwab <>
Submitted on : Thursday, June 16, 2016 - 3:24:05 PM
Last modification on : Tuesday, February 12, 2019 - 1:31:27 AM

File

extensions-def.pdf
Files produced by the author(s)

Identifiers

  • HAL Id : hal-01332850, version 1

Collections

Citation

Loïc Vial, Andon Tchechmedjiev, Didier Schwab. Extension lexicale de définitions grâce à des corpus annotés en sens. 23ème Conférence sur le Traitement Automatique des Langues Naturelles, Jul 2016, Paris, France. ⟨hal-01332850⟩

Share

Metrics

Record views

258

Files downloads

245