Daniel@FinTOC-2019 Shared Task : TOC Extraction and Title Detection - Archive ouverte HAL
Communication Dans Un Congrès Année : 2019

Daniel@FinTOC-2019 Shared Task : TOC Extraction and Title Detection

Résumé

We present different methods for the two tasks of the 2019 FinTOC challenge: Title Detection and Table of Contents Extraction. For the Title Detection task we present different approaches using various features : visual characteristics , punctuation density and character n-grams. Our best approach achieved an official F-measure score of 94.88%, ranking 6 on this task. For the TOC extraction task, we presented a method combining visual characteristics of the document layout. With this method we ranked first on this task with 42.72%.
Fichier principal
Vignette du fichier
article-Giguet-Lejeune.pdf (125.45 Ko) Télécharger le fichier
Origine Fichiers produits par l'(les) auteur(s)
Loading...

Dates et versions

hal-02303131 , version 1 (02-10-2019)

Identifiants

  • HAL Id : hal-02303131 , version 1

Citer

Emmanuel Giguet, Gaël Lejeune. Daniel@FinTOC-2019 Shared Task : TOC Extraction and Title Detection. The Second Financial Narrative Processing Workshop (FNP 2019), Sep 2019, Turku, Finland. pp.63-68. ⟨hal-02303131⟩
235 Consultations
752 Téléchargements

Partager

More