Email in French, Natural Language Processing and Pêle-mél project - Archive ouverte HAL Accéder directement au contenu
Communication Dans Un Congrès Année : 2023

Email in French, Natural Language Processing and Pêle-mél project

Résumé

The Pêle-Mél project, supported by the French Ministry of Culture and led by the University of Angers, has tested strategies for exploring corpora of French emails in order to promote their appraisial and exploitation. The Pêle-Mél project addressed the issue of electronic mailboxes from a resolutely interdisciplinary perspective, combining expertise in artificial intelligence (Leria laboratory) and archival science (Temos laboratory), and in collaboration with the archives mission of the Ministry of Health and the École des Chartes. Pêle-mél’s objective is to help the archivist to interpret and contextualize an individual mailbox or several mailboxes by categorizing and classifying the mails. To do this, it has developed a method that involves the extraction of named entities, mainly individuals and companies, the extraction of terms, the classification of terms and messages via word embeddings and document embeddings, i.e. by mobilizing a number of tools and techniques of natural language processing. The method is based on supervised machine learning and uses neural network models. It takes into account the linguistic specificities of French. The program has developed two prototypes, one to classify, the other to query metadata and explore content. Although the experimentation was based on a specific field – the Ministry of Health – the results are generalizable to other environments.
diapos_emailsymposium_20230614.pdf (3.39 Mo) Télécharger le fichier
Origine : Fichiers produits par l'(les) auteur(s)

Dates et versions

hal-04136335 , version 1 (21-06-2023)

Identifiants

  • HAL Id : hal-04136335 , version 1

Citer

Touria Ait El Mekki, Bénédicte Grailles. Email in French, Natural Language Processing and Pêle-mél project. Email archiving symposium, University of Illinois Urbana-Champaign, Jun 2023, Urbana (EU), United States. ⟨hal-04136335⟩
21 Consultations
2 Téléchargements

Partager

Gmail Facebook X LinkedIn More