Project 4β Impact: How Le French News Lab is Mapping Media Framing Across the French Press

Narmin Mammadova

2026-09-15

3 min read

AI Summary:

Supported by Oxylabs' pro bono initiative Project 4β, Le French News Lab built an automated pipeline that ingests roughly 1,900 articles daily across up to 30 national news outlets. By analyzing a four-year corpus of over 900,000 headlines, the Lab mapped the French press on a continuous ideology scale and revealed a measurable rise in loaded wording across coverage.

Overview

Le French News Lab is an independent research lab based in Paris, founded in 2025. It builds transparent and reproducible instruments for measuring media ecosystems and how news shapes public discourse. As a partner of Oxylabs' pro bono initiative Project 4β, the Lab has spent the past year measuring how the French press frames the news it covers.

Using Oxylabs' Web Scraper API provided through Project 4β, the Lab built a data-collection pipeline that ingests roughly 1,900 articles a day from up to 30 national French news outlets. Two papers built on it have been peer-reviewed and accepted for oral presentation at ICNLSP 2026 in Trento.

The challenge

Small teams cannot manually track an entire national press over years. It requires gathering consistent, large-scale public web data to move beyond anecdotal evidence to reliable insights. The methodological challenge lies in differentiating editorial story selection from specific wording choices. Analyzing hundreds of thousands of headlines over four years necessitated a scalable, consistent data collection method to accurately compare framing across outlets.

In practice, it means reading an entire national press across years of coverage, and telling apart two things a single bias score cannot separate: the stories an outlet chooses to put in front of you, and the words it chooses to describe them. That requires consistent, large-scale collection of public web data across dozens of outlets with different site structures and publishing rhythms. It is an infrastructure problem before it is a research problem, and it is where most small research teams stop.

Framing by Wording, Framing by Selection

Framing by Wording, Framing by Selection audited 902,111 deduplicated headlines from 25 national outlets between 2022 and 2025; to the Lab's knowledge, the largest framing-focused audit of French headlines to date. The study separated two decisions that most bias measurements collapse into one: whether an outlet chooses to run a story, and how it chooses to word it.

Holding the mix of stories constant, the share of headlines carrying a loaded wording device rose from 31.8% to 37.4% over four years. The stories the French press chose to cover barely changed. The language it used to tell them did.

Frldéo

FrIdéo is the first continuous ideology scale for the French press, placing 30 outlets on a single axis running from L'Humanité to Valeurs actuelles.

Rather than rely on a single proxy, the study fused nine independent families of evidence and found they converged far more than chance alone would explain, with a mean pairwise correlation of r = 0.84 across the best-covered families.

Because the scale reads current articles rather than static labels, it also catches editorial change as it happens.

Key Findings

Metric	Result
Headline corpus	902,111 deduplicated headlines, 25 national outlets, 2022–2025
Ideology scale	30 outlets on one continuous scale, from nine independent families of evidence
Source agreement	Mean pairwise correlation of r = 0.84 across the best-covered evidence families
Wording device rate	Rose from 31.8% to 37.4% over four years, story mix held constant
Classification benchmark	87,637 labeled articles across 13 publishers
Role-framing study	28,592 headlines comparing coverage of two major parties
Daily ingest	About 1,900 articles a day

Why Public Web Data Matters

Understanding how a national press frames the news requires more than a small sample. It depends on following patterns over years and comparing outlets on the same footing, which is only possible with reliable access to public web data at scale. Consistency is what makes that comparison legitimate, and a gap in collection is not a missing row in a spreadsheet: it is a stretch of time in which one outlet's language cannot be compared with anyone else's.

That is what Project 4β is built to provide: professional-grade collection infrastructure for researchers on limited budgets, so their time goes into analysis, publication and impact rather than into keeping scrapers alive.

Every claim we publish about the French press rests on the same thing: that the same outlets were collected the same way, across four years of coverage, with no gaps. That consistency is the instrument. Project 4β is what lets a lab our size maintain it across thirty publishers and still spend its time on the research.

Amr Sobhy, Founder and Director of Le French News Lab

Through Project 4β, Oxylabs supports researchers working to understand the information ecosystem and turn public web data into meaningful analysis. To learn more, contact the team via the form or at 4beta@oxylabs.io.

Forget about complex web scraping processes

Choose Oxylabs' advanced web intelligence collection solutions to gather real-time public data hassle-free.

About the author

Narmin Mammadova

PR Content Manager

Narmin is the PR Content Manager for Project 4β at Oxylabs. She enjoys the challenge of getting people to care, and pro bono work gives her good stories to tell. In her spare time, she travels whenever possible or indulges her love of poetry and reciting.

Learn more about the author Narmin Mammadova Learn more about the author Narmin Mammadova

All information on Oxylabs Blog is provided on an "as is" basis and for informational purposes only. We make no representation and disclaim all liability with respect to your use of any information contained on Oxylabs Blog or any third-party websites that may be linked therein. Before engaging in scraping activities of any kind you should consult your legal advisors and carefully read the particular website's terms of service or receive a scraping license.