Research

How do continuous news channels decide what counts as news, and who gets to speak about it?

I work on the sociology of media and journalism with the methods of computational social science: large broadcast corpora, built and documented so that each measurement can be traced and reproduced.

The Laudisi Observatory

A pipeline I designed and maintain to record, transcribe and analyse French continuous news channels: stream capture, automatic speech recognition, speaker diarisation and OCR of on-screen banners.

Channels
BFMTV, CNews, franceinfo and LCI: on-screen banners on all four, speech on the first three
First corpus
March 2026, about 2,230 hours of broadcast
Methods
Topic segmentation, unsupervised clustering, sentence embeddings
Code
Project description on GitHub

The project examines agenda and framing dynamics across channels: how far they converge in the topics they cover and the words they use, which channel takes up a topic first, and who gets to speak on air, including the gap between how often women are seen and how often they are heard.

Theses

  1. In progress

    Master’s (M2) thesis on the mediated construction of socially relevant reality on continuous news channels: professional routines, framing structures and news formats. Supervised by Marshall A. Taylor (ENSAE-CREST).

  2. Mar 2026

    L’Observatoire Laudisi : sociologie computationnelle de l’information en continu française. Master’s (M1) thesis, supervised by Samuel Bouron and Antoine Mazières (IRISSO). Prix de la Recherche Impact Tank × L’Ascenseur.

Earlier projects

  1. 2025

    Farming channels on YouTube: digital ethnography of farmer vloggers, brands and aesthetic creators, and text analysis of about two million comments. LISIS, with Baptiste Kotras and Sylvain Brunier.

  2. 2022–25

    Bachelor’s research projects: the colonial dimension of knowledge relations in the chlordecone scandal in Martinique and Guadeloupe; the role of soil in urban development at Porte de Montreuil, Paris.

Talks

  1. AI & NLP Workshop, ACSS-PSL Institute (online).

  2. Montrées plus qu’entendues ? Mesurer en continu l’écart entre présence à l’écran et accès à la parole des femmes sur les chaînes d’information françaises

    Study day « Mesurer la diversité de genre dans les contenus médiatiques : l’apport des sciences sociales computationnelles », Grenoble.

  3. The Laudisi Observatory: A computational pipeline for framing and agenda dynamics in French continuous news

    Computational Humanities Research Group Seminar Series, King’s College London (online).