With Etienne Ollion and Guilhem Sicard, I have developed an original methodology that leverages supervised machine learning and BERT models to automatically extract quotes – i.e., reported speech triplets composed of a source, a cue verb, and a statement – from large-scale French and English news corpora. We have presented QED in several occasions, and we are currently finalizing a paper documenting this methodology and preparing a GitHub repository. Stay tuned for more updates!
In my dissertation, I explore how journalists include political discourse in news narratives through the act of quoting political actors - put otherwise, how political discourse structures news narratives. I combine large-scale computational analysis of over 2 million quotes of politicians extracted from French newspapers (1995-2025) with 35 interviews with political journalists to draw a historical portrait of quoting practices in French legacy journalism and to analyze the stakes of quotation in a changing media and political environment. I am currently working on a paper where I analyze how quote patterns in French national newspapers for different party families evolve as a function of these parties’ electoral trajectories and presence in French political institutions.
Emma Bonutti D'Agostini, PhD Candidate in Sociology
Last update: July 2026.