Skip to Main content Skip to Navigation
Conference papers

Indexation et appariements de documents cliniques pour le Deft 2019

Abstract : Indexing and pairing texts of the medical domain In this paper, we propose different methods for the indexation and pairing tasks of the 2019 edition of the french text mining challenge (Deft). For indexing texts we proposed two different methods, one exploit a pairing method developed for task 2 in order to use keywords found in similar documents.. The second method takes advantage of terminological annotation obtained via the MeSH (Medical Subject Headings) resource. For the pairing task, we also proposed two methods. The first one is unsupervised and relies on similarities at character-level, in the second one we exploited siamese networks. In both tasks our results were quite disappointing except for task 2 where the unsupervised method achieved an honourable 62% score.
Complete list of metadatas

Cited literature [6 references]  Display  Hide  Download

https://hal.archives-ouvertes.fr/hal-02467526
Contributor : Gaël Lejeune <>
Submitted on : Wednesday, February 5, 2020 - 9:23:27 AM
Last modification on : Monday, March 2, 2020 - 6:24:48 PM

File

Buscaldi.pdf
Publisher files allowed on an open archive

Identifiers

  • HAL Id : hal-02467526, version 1

Citation

Davide Buscaldi, Dhaou Ghoul, Joseph Le Roux, Gaël Lejeune. Indexation et appariements de documents cliniques pour le Deft 2019. TALN -Défi Fouille de Textes (DEFT) 2019, Jul 2019, Toulouse, France. ⟨hal-02467526⟩

Share

Metrics

Record views

27

Files downloads

7