Methodology behind Framing Gaza
A deep dive into how we applied Association Rule Mining to over 25,000 mass-media articles covering the events in Gaza.
Overview
Framing Gaza was built to test a concrete question: can large-scale text analysis document and measure recurring patterns in how major Western outlets narrate the war in Gaza? The methodology matters because it moves the conversation about framing away from anecdotal examples and toward a reproducible, quantifiable corpus-level analysis.
Data sources
The dataset brings together more than 25,000 English-language articles from 16 news outlets in the United States, the United Kingdom, France, and Germany. The corpus focuses on digital text coverage and reflects the material we were able to collect and normalize from those outlets during the study period.
Method
We cleaned and standardized article text, then applied Association Rule Mining (ARM) with the Apriori algorithm to detect which terms and concepts repeatedly appeared together, and to measure the strength and consistency of those associations. We used the resulting recurring associations, weighted by their statistical strength, as a way to surface dominant framing patterns that can then be interpreted alongside close reading and newsroom context.
Limitations & caveats
ARM detects and measures recurring co-occurrence patterns; it does not, on its own, prove intent, causality, or editorial coordination. The findings are limited to the sampled outlets, the collection window, and the material available online, so they should be read as evidence of recurring narrative structure rather than an exhaustive account of all coverage.
Why this methodology matters
This workflow gives journalists and researchers a repeatable, measurable way to investigate how narrative patterns consolidate across outlets at scale. It is most useful when paired with qualitative reporting that can test which associations signal durable framing and which reflect short-lived event cycles.