Research Projects

Submitted for peer review or ongoing research projects.

Submitted

Beyond Evidence: How Framing Shapes Public Health Policies During Health Crises

Article submitted to Policy Studies Journal (R&R)

This study investigates how framing, evidence, and the roles of scientists and political decisionmakers influence public health policy decisions during the COVID-19 pandemic in Quebec and Sweden using NLP analysis.

Résumé

This study investigates how framing, evidence, and the roles of scientists and political decisionmakers in policymaking influence public health policy decisions during the COVID-19 pandemic in Quebec and Sweden. Utilizing a comprehensive dataset of press conference transcripts, we apply natural language processing (NLP) to assess the impact of different framings on suppression and mitigation policies. Our analysis reveals that framing affects policy decisions, often independent of evidence. In Quebec, where political decisionmakers were central, a Dangerous framing, which emphasizes the severe health threats of COVID-19, is associated with an increase in stringent suppression policies, even in the absence of strong evidence.

COVID-19FramingPublic PolicyNLPPublic HealthComparative Analysis

Document complet

Télécharger PDF

Authors

Antoine LemorÉric Montpetit

Far-Right Ideas in the National Assembly? A Computational Study on the Prevalence of Far-Right Ideology in the French Prime Ministers' Policy Speech Since 1959

Article accepted in Revue française de science politique (2025)

This study analyzes the diffusion of far-right ideas in French Prime Ministers' general policy statements (1959-2024) using Natural Language Processing (NLP) methods.

Résumé

This study analyzes the diffusion of far-right ideas in the general policy statements of French Prime Ministers (1959–2024) using Natural Language Processing (NLP) methods. It develops and introduces the Far-Right Ideological Score (SIED), a quantitative indicator designed to measure the proportion of far-right ideas within political discourse. The results highlight three significant periods: a peak during the Algerian War (1959–1961), a decline following May 1968, and a steady rise from the 1970s onward, intensifying after 2005. The study reveals a cross-party diffusion of far-right ideas, particularly driven by center and right-wing forces.

Far-rightDiscourse analysisNLPFrench politicsMetapoliticsCo-optation

Document complet

Télécharger PDF

Authors

Tristan BoursierAntoine Lemor

Ongoing

When climate science speaks, do policymakers respond? A computational analysis

CCF Project - Canadian Climate Framing

This study uses a machine-learning-annotated corpus of over 250,000 Canadian newspaper articles (1988–present) to examine whether policymakers' climate-related media interventions follow scientific appeals.

Résumé

This study uses a machine-learning-annotated corpus of over 250,000 Canadian newspaper articles (1988–present) to examine whether policymakers' climate-related media interventions follow scientific appeals. As part of the CCF-Canadian-Climate-Framing Project led by Alizée Pillod, this research investigates the temporal dynamics between scientific discourse and political responses in the Canadian context of climate change. Using advanced NLP techniques, we analyze patterns of responsiveness, lag times, and the evolution of climate framing in both scientific and political spheres.

Climate changeComputational analysisPolitical discourseScience communicationCanadian mediaNLP

Document complet

Télécharger PDF

Authors

Antoine LemorAlizée PillodMatthew Taylor

YOUPOL: A Textual Database of Over 20,000 Videos by Francophone Political Influencers on YouTube (2006–2024)

Analysis of Online Political Radicalization

Computational study of discourse and ideas from francophone political influencers on YouTube, analyzing over 20,000 automatically transcribed videos.

Résumé

This project develops YOUPOL, a unique textual database containing automatic transcriptions of over 20,000 videos from francophone political influencers on YouTube from 2006 to 2024. Using advanced natural language processing and machine learning techniques, we analyze the evolution of online political discourse, radicalization dynamics, and influence networks. This resource enables the study of how political ideas propagate, transform, and polarize within the francophone digital ecosystem.

YouTubePolitical radicalizationInfluencersNLPDatabaseDiscourse analysisFrancophone politics

Document complet

Télécharger PDF

Authors

Antoine LemorTristan Boursier

Credible Science, Influential Science? A Computational Study on the Influence and Credibility of Public Scientific Research Agencies in Public Health in Canada

Analysis of Epistemic Authority of Scientific Institutions

Computational study on the influence of public scientific research and advisory agencies in public health in Canada, analyzing their credibility and impact on public policies.

Résumé

This study examines the relationship between scientific credibility and political influence of public health research agencies in Canada. Using advanced computational methods, we analyze how the epistemic authority of these institutions translates (or fails to translate) into influence on public policy decisions. The study focuses on several key agencies, including INSPQ, Health Canada, and the Public Health Agency of Canada, examining their role during recent health crises. The results reveal complex dynamics between scientific expertise, institutional legitimacy, and political influence.

Public healthScientific advisoryEpistemic authorityPublic policyCanadaINSPQComputational analysis

Document complet

Télécharger PDF

Authors

Antoine Lemor

Building Lasting Trust: Analyzing the Resilience of Science-to-Policy Institutions in Quebec and Sweden During the COVID-19 Pandemic

ENDURE Project - Institutional Resilience

This project compares the resilience of science-to-policy institutions in Quebec and Sweden during the COVID-19 pandemic, focusing on their ability to maintain public trust.

Résumé

As part of the ENDURE project, this comparative study examines the resilience of science-to-policy institutions in public health in Quebec and Sweden during the COVID-19 pandemic. By analyzing communication strategies, organizational structures, and adaptation mechanisms, we assess how these institutions maintained (or lost) public trust in the face of the unprecedented challenges of the pandemic. The study reveals contrasting approaches: the centralized Quebec model versus the decentralized Swedish model, each presenting distinct strengths and vulnerabilities in terms of institutional resilience.

COVID-19ResiliencePublic trustScientific institutionsQuebecSwedenPublic healthENDURE project

Document complet

Télécharger PDF

Authors

Antoine LemorPhilippe Bourbeau

LLM Tool: A Hybrid Pipeline for Automated Large-Scale Text Annotation Using Local Language Models and BERT Classifiers

Development of a Novel Methodological Approach for Automated Annotation

This project presents LLM Tool, an innovative processing pipeline that combines locally-run Large Language Models (LLMs) with BERT classifiers to enable fully automated text corpus annotation at scale. This approach revolutionizes text annotation in computational social sciences.

Résumé

The annotation of large-scale text corpora represents a fundamental bottleneck in computational social science research, particularly when dealing with complex multi-label classification tasks in political science. LLM Tool is a novel hybrid pipeline that combines local Large Language Models (Gemma3:27B, Llama3.3:42B, Nemotron:42B, DeepSeek-R1:70B, GPT-OSS:120B) with BERT-based classifiers to enable fully automated annotation at scale. Our approach leverages these state-of-the-art open-source LLMs, running entirely on local infrastructure, to generate initial annotations on stratified samples, which then serve as training data for specialized BERT models capable of efficient large-scale inference. The pipeline implements an extended version of the Comparative Agendas Project (CAP) coding scheme adapted for Canadian political discourse, generating structured annotations across 21 policy themes, 9 political parties, 2 specific themes, and 3 sentiment dimensions. Empirical validation on 1,593 manually annotated Canadian parliamentary debates and media articles demonstrates that BERT models trained on LLM-generated annotations achieve a Micro F1 score of 0.6673, significantly outperforming models trained on human annotations (0.4601)—a 45% improvement.

Artificial IntelligenceLLMBERTAutomated AnnotationComputational Social SciencesMulti-label ClassificationCanadian PoliticsNLPMachine Learning

Document complet

Télécharger PDF

Authors

Antoine LemorJérémy GilbertShannon DinanYannick Dufresne