Correction automatique d'attachements prépositionnels par utilisation de traits visuels

Abstract : Resolving prepositional attachments is a syntactic task that requires semantic knowledge, which can be extracted from the composition of a picture associated with a text. We present and analyse the difficulties of performing this task, for which we build a full system trained on extended annotations from Flickr30k Entities. When lexical semantics are unavailable, visual information brings 3% improvement.
Complete list of metadatas

Cited literature [24 references]  Display  Hide  Download

https://hal.archives-ouvertes.fr/hal-01832979
Contributor : Sebastien Delecraz <>
Submitted on : Monday, July 9, 2018 - 11:34:02 AM
Last modification on : Tuesday, April 16, 2019 - 1:41:19 AM
Long-term archiving on : Tuesday, October 2, 2018 - 8:16:21 AM

File

taln-2018.pdf
Files produced by the author(s)

Identifiers

  • HAL Id : hal-01832979, version 1

Citation

Sebastien Delecraz, Leonor Becerra-Bonache, Benoit Favre, Alexis Nasr, Frédéric Béchet. Correction automatique d'attachements prépositionnels par utilisation de traits visuels. 25ème conférence sur le Traitement Automatique des Langues Naturelles, May 2018, Rennes, France. ⟨hal-01832979⟩

Share

Metrics

Record views

106

Files downloads

88