Research Engineer in Digital Methods

Current Work

I work at the French médialab, an interdisciplinary research laboratory that investigates the role of digital technology in our societies. My research focuses on the circulation of information between social networks and traditional media. I use methods from data mining, natural language processing and computer vision. I contribute to the development of open-source software designed for research in humanities and social sciences.

Open-Source Software

pimmi: clustering of images based on total or partial copy detection
twembeddings: my Python scripts using LLM embeddings to detect events in a stream of documents (tweets, news, etc.)
stories: a Rust implementation of the event detection algorithm detailed in twembeddings
gazouilloire: long term tweets collection

PhD

During my PhD, I worked on mutual influences between social media and traditional news media. My research project was bi-disciplinary: to make sense of datasets containing millions of messages a day, I used machine learning techniques. On the other hand, I used econometrical methods to understand the causal effect of social media activity on publications by news media. My thesis was funded by the Institut National de l’Audiovisuel (the French Multimedia Institute) and co-supervised by CentraleSupélec.

Main supervisor:

Céline Hudelot CentraleSupelec, Mathematics and Interaction with Computer Sciences (MICS) Laboratory

Co-supervisors:

Julia Cagé Sciences Po, Department of Economics

Nicolas Hervé Institut National de l’Audiovisuel, Research Department

Publications

Créac’h, M., & Mazoyer, B. (2024). “Tableau vivant memes in lockdown: virality and vitality.” Hybrid. Revue des arts et médiations humaines.

Ray, M., Wang, Q., Mélanie-Becquet, F., Poibeau, T., & Mazoyer, B. (2024). “An Incremental Clustering Baseline for Event Detection on Twitter”. In Proceedings of the Workshop on the Future of Event Detection (FuturED), 18-24.

Mazoyer, B., Hervé, N., Hudelot, C., & Cagé, J. (2024). “Comparison of Short-Text Embeddings for Unsupervised Event Detection in a Stream of Tweets”. In Advances in Knowledge Discovery and Management, Volume 10, 81–97 · preprint

Cagé, J., Hervé, N. & Mazoyer, B. “Social Media and Newsroom Production Decisions”. Submitted.

Mazoyer, B., Cagé, J., Hervé, N. & Hudelot, C. (2020). “A French Corpus for Event Detection on Twitter”. In “International Conference on Language Resources and Evaluation (LREC 2020)”, 6220–6227

Evrard, M., Uro, R., Hervé, N. & Mazoyer, B. (2020). “French Tweet Corpus for Automatic Stance Detection”. In “International Conference on Language Resources and Evaluation (LREC 2020)”, 6317–6322

Mazoyer, B., Hervé, N., Hudelot, C., & Cagé, J. (2020). “Représentations lexicales pour la détection non supervisée d’événements dans un flux de tweets : étude sur des corpus français et anglais”. In “Extraction et Gestion des Connaissances (EGC 2020)”

Mazoyer, B., Cagé, J., Hudelot, C., & Viaud, M.-L. (2018). “Real-Time Collection of Reliable and Representative Tweets Datasets Related to News Events”. In “Proceedings of the First International Workshop on Analysis of Broad Dynamic Topics over Social Media (BroDyn 2018) co-located with the 40th European Conference on Information Retrieval (ECIR 2018)”, 23–34

Mazoyer, B., Turenne, N., & Viaud, M.-L. (2017). “Étude des influences réciproques entre médias sociaux et médias traditionnels”. In “Amsaleg, L., Claveau, V. & Tannier, X. Actes de l’atelier Journalisme Computationnel 2017”, 37–40

Communications and Posters

DigitHum 2023 Mazoyer, B. et Plique, G. “Explorer des réseaux dans un carnet interactif avec ipysigma”. @ “Workshop Digit_Hum 2023: Manipuler des données en SHS, R, Python et les carnets interactifs” · slides · video

FOSDEM 2023 Hervé N. et Mazoyer B. “PIMMI: a command line interface to study image propagation” · slides · video

“Automatically detect and archive media events on Twitter” Médialab Research Seminar, December 2019, Sciences Po Paris, France.

Mazoyer, B., Hervé, N., Hudelot, C., & Cagé, J. (2019). “Réduire les biais dans la collecte de tweets”. In “Journée DAHLIA - Informatique et Humanités numériques : quelles problématiques pour quels domaines ?”, June 24, 2019, Nantes, France.

Mazoyer, B., (2019). “Using Text and Image for Topic Detection on Twitter”. Poster at the Workshop on Representation Learning for Complex Data, May 24, 2019, Lyon, France.

“Capter les tweets liés à l’actualité” PEPS EXIA Seminar, October 2016, Marne-la-Vallée, France