Yann Guérin

To acquire the data from these forums we use web scraping. I wrote a scraping script which automates the process of collecting all the posts on the forum along with associated thread and user level metadata. The scripts run every morning to collect new posts.   The forums contain several million posts across hundreds of thousands of threads written by tens of thousands of users. All the text and metadata from these forums is stored in MongoDB, a noSQL database well suited for the large amounts of text data we work with.

I built a web app for exploring the data collected from scraping and stored in MongoDB. Since we work with sociologists who aren't necessarily knowledgeable in noSQL queries we wanted a web app which provides a no-code platform for exploring the data.   This web app was built using Plotly Dash. I used Dash since it is a versatile framework, built on top of Flask, with lots of flexibility and ability to work with large amounts of data. The web app connects to MongoDB and queries the data from there.

To characterize the radical ideas being discussed on the forums we use topic models. These models have been around for decades, and over time multiple kinds of topic models have been created. We have explored and tested several on our data.   The two main types of models we have explored are LDA and BERTopic.

Beyond characterizing the radical ideas being discussed on these forums, we also seek to understand how they are shared and spread between users. To analyze this we turn to social network analysis and information diffusion.

I was responsible for building the model used to explore the relationship between luck and circumstance. With the guidance of Dr. Yukiko Asada and others on the team I built the model to incorporate the two main components: random luck and socioeconomic circumstances.

I had the opportunity to present the microsimulation work to the 2025 Politics, Philosophy, and Economics conference in London, UK.   I developed the slides for the presentation. I focused on making the slides and the presentation understandable to the multidisciplinary audience of the conference. I gave a 20 minute presentation on my work building the microsimulation model.

I co-authored two research papers describing this work. I contributed to writing the methods section and I created the figures used in the papers. One paper was submitted to Social Indicators Research and is currently under review. The second is actively being developed.

The goal for this project was to analyse the trends across all OBGYN papers in PubMed. In total, PubMed has around 27 million papers, so the first problem was filtering that down to the papers which were likely OBGYN related.   We developed a search query for PubMed that included a wide range of OBGYN terms which filtered the papers down to 4.8 million.

Filtering¶ I created a controlled vocabulary from a subset of MESH terms which were related to OBGYN.   I processed the paper titles and abstracts using MetaMap in a Python script to identify the the concepts and terms in this controlled vocabulary. Extra conditions were added for some terms that were ambiguous.   Using the concepts extracted and the controlled vocabulary I filtered down the 4.8 million papers to 955,626. Classifying the Papers¶ We were interested in the trends in gendered language so three classes were created: gendered, gender neutral, and transgender. The assignment of m...

I had the opportunity to present my work on this project to the Dalhousie Department of Obstetrics and Gynaecology Research Day Conference in 2025.  

I am first-author on a manuscript we published to the Journal of Obstetrics and Gynaecology Canada in February 2026. I wrote the methods section, created the figures, and created the supplementary document for this paper. I collaborated with Professor Jocelyn Stairs, Professor Finlay Maguire, and Dr. Aisling Clancy on this paper.   Beyond “Women’s” Health: A Natural Language Processing Analysis of Trends in Gender-Neutral Language Use