Deep Learning Models and Tools (ModOAP)

Deep Learning Models and Tools (ModOAP)

The ModOAP project is designed around two primary objectives. Firstly, we seek to develop an analytic toolbox for large databases to be used in social and human sciences research. This will be based on models that can be easily adapted to different types of corpuses of text and images, and will include programmes that can be reused by the institutions and research laboratories involved in Labex, and the broader SHS scientific community. These tools (classification of images, textual units, recognition of structures, similarity detection, etc.) will be the object of a collective development in partnership with other Labex projects and training workshops aimed at researchers. Secondly, the project aims to test these tools on two corpuses that are particularly important for the analysis of the construction of collective memory – analysis of school textbooks, through the digital reserves of the BNF, and of photo-essays through the Kagan collection kept at La Contemporaine library-museum. Using analysis of large databases enabled by deep learning tools, we aim to explore the construction of institutional and academic memory and its transmission through textbooks, as well as looking at the way the press can contribute to shaping media-based collective memory through the re-use and re-publication of images over the long term.

Project Leader

Julien SCHUH , Université Paris Nanterre - CSLF EA 1586

Internal cluster partners

  • Bibliothèque nationale de France (BnF)
  • Centre des Sciences des Littératures en langue Française (CSLF) - EA 1586
  • La contemporaine | bibliothèque, archives, musée des mondes contemporains
  • Modèles, Dynamiques, Corpus (MoDyCo) - UMR 7114
  • IR* Huma-Num

Associated partners

Projet ANR "Numapresse"

Projet ARTEC "La preuve par l'image"

Projet IUF "Synthétismes fin de siècle"


24 months


apprentissage profond, intelligence artificielle, boîte à outils, humanités numériques