The Book Structure Extraction Competition with the Resurgence full content software at Caen University - Archive ouverte HAL Access content directly
Book Sections Year : 2012

The Book Structure Extraction Competition with the Resurgence full content software at Caen University

Abstract

The GREYC participated in the Structure Extraction Competition, part of the INEX/ICDAR Book track, for the third time, with the Resurgence software. We used a minimal strategy primarily based on full-content top-down document representation with two then three levels, part, chapter and section. The main idea is to use a model describing relationships for elements in the document structure. Frontiers between high-level units are detected. The periphery center relationship is calculated on the entire document and then reflected on each page. The weak points of the approach are that level hierarchy is implicit, and dependent on named levels. It does not fit with the chapter and section levels reflected in the ground-truth. The strong points are that it deals with the entire document; it handles books without ToCs, and extracts titles that are not represented in the ToC (e. g. preface); it is tolerant to OCR errors and language independent; it is simple and fast. A test on sections was run after the competition to help understand the evaluation issues with more than two levels.
Fichier principal
Vignette du fichier
CHS-GIGUET-2012-1.pdf (4.24 Mo) Télécharger le fichier
Origin : Files produced by the author(s)

Dates and versions

hal-01071717 , version 1 (13-10-2014)

Identifiers

Cite

Emmanuel Giguet, Nadine Lucas. The Book Structure Extraction Competition with the Resurgence full content software at Caen University. Focused Retrieval of Content and Structure, Springer, pp86-97, 2012, Lecture Notes in Computer Science, 978-3-642-35733-6. ⟨10.1007/978-3-642-35734-3_7⟩. ⟨hal-01071717⟩
150 View
90 Download

Altmetric

Share

Gmail Facebook X LinkedIn More