Project Incendiary
A study of how tone in major news headlines shifted across 15 years, built on 4.5 million archived headlines.
- Role
- TODO: role
- When
- August–October 2023
- Made with
- Python, Selenium, RoBERTa, Hugging Face Transformers, Pandas, Chart.js, SvelteKit
Problem
I wanted to know whether headlines from major outlets had grown more negative or inflammatory since the 2016 election.
What I did
I scraped 4.5 million headlines from the Wayback Machine, trained a bias detector on expert annotations plus my own, and scored every headline.
TODO: choose consistent framing: media bias, positivity, or inflammatory headlines. Add one results chart.
What changed
TODO: the headline finding, in one sentence with its number.
TODO: confirm whether the Congressional Award, NY-06 recognition was a finalist placement or an award before adding that claim.