French continuous news channels broadcast around the clock, yet the discursive flow they produce has remained difficult to study systematically. This talk presents the Laudisi Observatory, a computational pipeline built to record, transcribe and analyse the full audiovisual output of France’s three main rolling-news channels — BFMTV, Cnews and France Info — across a complete month (March 2026, roughly 2,230 hours of broadcast).Rather than testing for already-documented political biases, the project asks how this ecosystem works: what subject got the authorization to be on screen, how the channels talk about it, who gets a voice on screen and what is left unsaid. The pipeline pairs sociological theories of news production with established computational methods. Topics are segmented endogenously from the transcriptions — using TextTiling and HDBSCAN— so that the corpus defines the relevant units. Lexical specificity is measured with Monroe log-odds z-scores and framing divergence between channels with Jensen-Shannon divergence. The talk walks through the architecture and the methodological decisions it required, then turns to what the coexistence of competing frames suggests about the fragmentation of a shared public sphere.

Speaker:

Yann de Boisvilliers is a Master's student in Économies et Sociétés at Paris-Dauphine and EHESS (PSL University, Paris). He holds a BA in Sciences pour un Monde Durable (Economics and Social Sciences) from PSL University (2022–2025). His work sits at the intersection of media sociology and computational social science. His thesis examines how French rolling-news channels — through professional routines, framing structures and informational formats — participate in the mediated construction of socially relevant reality. He is a research assistant at the Institut des Sciences Sociales du Politique (ENS Paris-Saclay / LSE), where his master's research is supervised by Baptiste Coulmont and Benoît de Courson. He spent June 2026 at the London School of Economics as a visiting researcher, and is more broadly interested in how computational methods can illuminate transformations in public language and collective attention.