TCThibault Clérice

Thibault Clérice

*This is an algorithmically generated author record based on publicly available information.

Current Organizations / Institutes

The organizations/institutes from the scholar's latest article

Previous Organizations / Institutes

The organizations / institutes from the scholar's other articles in the past ten years


Thibault Clérice has a total of 32 co-authors. Here are the top co-authors by number of shared publications:


Latest Publications

  • Structure-Aware Text Recognition for Ancient Greek Critical Editions

    by Nicolas Angleraud, Antonia Karamolegkou, Benoît Sagot, Thibault Clérice

  • Détection d'isotopies par apprentissage profond : l'exemple de la sexualité en latin classique et tardif

    by Thibault Clérice

    En 1982, James N. Adams a produit l’étude de référence sur le vocabulaire latin de la sexualité. Dans son ouvrage, l’auteur présente un large éventail de mots ou d’expressions classées en fonction de leurs caractéristiques stylistiques (métaphore, métonymie, etc.) et sémantiques (guerre, cuisine, chasse, violence...) dans diverses catégories (actes, parties intimes masculines et féminines, etc.). À la même époque, François Rastier redéfinit l’isotopie comme « la récurrence d’un même trait sémantique » dans un texte. Dans la lignée de ces travaux, nous avons construit un méta-corpus latin inédit de 20 millions de mots en TEI mêlant projets existants et nouveaux textes. En utilisant ce métacorpus et le travail d’Adams, nous avons construit un exemplier numérique qui fournit 2500 exemples d’isotopies sexuelles latines. Pour traiter le corpus, nous avons développé une méthode et les outils pour lemmatiser et annoter la morphosyntaxe des textes en latins classique et tardif. Enfin, nous avons testé des méthodes d’apprentissage profond pour détecter les isotopies dans des textes latins allant de 200 avant notre ère jusqu’à +700. Cette méthode pourra servir de base à la construction de nouveaux « vocabulaires » ou exempliers pour d’autres isotopies à l’avenir. Une sélection de méthodes montre des résultats robustes avec l’exemplier complet, et nous évaluons les limites des diverses architectures de modèle de détection d’isotopie en fonction de la taille de l’exemplier ou de la part d’implicite dans ce dernier.

  • Open Access

    Wauchier, Is That You? A Multi-Manuscript Authorship Analysis of Saint Lambert’s Life

    by Thibault Clérice, Ariane Pinche

    We investigate the possible authorship of the Vie de saint Lambert, an anonymous Old French hagiographic text. Building on previous research that noted a stylistic proximity between this text and the Seint Confessor attributed to Wauchier de Denain, we expand the inquiry across ten manuscripts. Alongside our computational analysis, we provide a brief study of the manuscript tradition and the transmission of the Vita of Saint Lambert within the broader literary genre of hagiography. We work directly with automatic transcriptions from medieval manuscripts, enhancing the tooling for post-processing space errors and lemmatization. We employ the recent Bootstrap Distance Impostors (BDI) approach to improve precision in authorship verification tasks, and propose a methodology for addressing the challenges of massively anonymous corpora in this context. Our results confirm a tendency to attribute the Vie de saint Lambert to Wauchier de Denain, though the stylistic alignment is less marked than in texts firmly attributed to him.


  • Open Access

    “I Am Too Old For This Style!” A Stylometric Benchmark of Age Effect on Authorship Attribution

    by Florian Cafiero, Lucence Ing, Simon Gabay, Thibault Clérice

    Can age act as a significant factor in determining authorship? While writers’ styles are known to measurably evolve over time, most computational pipelines still treat them as time-invariant. Building on recent work that tracks idiolectal change in several literary traditions, we introduce the first controlled benchmark that quantifies how both the magnitude and the direction of temporal distance affect attribution. A diachronic corpus of French novelists is sampled at bidirectional gaps of ±1, ±5, ±10, and ±15 years. Verification is then tested with a generative Bootstrap Distance Impostors (BDI) model and a discriminative linear Support Vector Machine (SVM). Results reveal a near-linear loss of confidence as the gap widens, but –critically– also show a systematic directional asymmetry: late-career texts remain recognisable when compared with early-career references, whereas early texts are markedly harder to verify against late ones, suggesting a cumulative rather than a substitutive evolution. The pattern persists under random-pair controls, confirming that it is not a sampling artefact. All code, data, and evaluation scripts are released to encourage further research on temporally robust authorship analysis and to quantify both the dynamics and amplitude of idiolectal change over time.


    • 18 July 2025
    • Published by HAL in HAL
    Open Access

    "I am too old for this style!" A stylometric benchmark of age effect on authorship attribution

    by Florian Cafiero, Lucence Ing, Simon Gabay, Thibault Clérice

    Is age a lethal weapon for authorship verification? While writers' style is known to measurably evolve, most computational pipelines still treat style as time-invariant. Building on recent work that tracks idiolectal drift in several traditions, we introduce the first controlled benchmark that quantifies how both the magnitude and the direction of temporal distance affect attribution. A diachronic corpus of French novelists is sampled at bidirectional gaps of ±1, ±5, ±10 and ±15 years. Verification is then tested with a generative Bootstrap Distance Impostors (BDI) model and a discriminative linear Support-Vector Machine (SVM). Results expose a near-linear loss of confidence as the gap widens, but -critically -also reveal a systematic directional asymmetry: late-career texts remain recognisable from early-career references, whereas early texts are markedly harder to verify against late ones, suggesting cumulative rather than substitutive stylistic change. The pattern persists under random-pair controls, confirming that it is not a sampling artefact. All code, data and evaluation scripts are released to encourage further research on temporally robust authorship analysis and to reassess the default assumption of a fixed stylistic "fingerprint.

  • Open Access

    Reconnaissance automatique d’écriture et sources historiques. Limites et nouvelles perspectives

    by Ariane Pinche, Thibault Clérice

    The aim of this article is not to present unpublished research, but to review the work we have been carrying out since 2019 at the École nationale des chartes, then at INRIA Paris and the CNRS, on Automatic Text Recognition (ATR), and more particularly on its application to medieval manuscripts. This article provides an overview of our research, together with references for exploring in detail specific aspects of our past and current work.


  • Top Cited Papers

    CATMuS Medieval: A Multilingual Large-Scale Cross-Century Dataset in Latin Script for Handwritten Text Recognition and Beyond

    by Thibault Clérice, Ariane Pinche, Malamatenia Vlachou-Efstathiou, Alix Chagué, Jean-Baptiste Camps, Matthias Gille Levenson, Olivier Brisville-Fertin, Federico Boschetti, Franz Fischer, Michael Gervers, et al.


  • Open Access

    You Actually Look Twice At it (YALTAi): using an object detection approach instead of region segmentation within the Kraken engine

    by Thibault Clérice

    Layout Analysis (the identification of zones and their classification) is the first step along line segmentation in Optical Character Recognition and similar tasks. The ability of identifying main body of text from marginal text or running titles makes the difference between extracting the work full text of a digitized book and noisy outputs. We show that most segmenters focus on pixel classification and that polygonization of this output has not been used as a target for the latest competition on historical document (ICDAR 2017 and onwards), despite being the focus in the early 2010s. We propose to shift, for efficiency, the task from a pixel classification-based polygonization to an object detection using isothetic rectangles. We compare the output of Kraken and YOLOv5 in terms of segmentation and show that the later severely outperforms the first on small datasets (1110 samples and below). We release two datasets for training and evaluation on historical documents as well as a new package, YALTAi, which injects YOLOv5 in the segmentation pipeline of Kraken 4.1.