Methodology

Restworld measures attention, not opinion.

The platform is built around rolling time windows so the output reflects what is happening now rather than a static historical average.

1. Data Ingestion and Rolling Windows

Articles are ingested from outlets across countries and regions. Each outlet is mapped to country and region metadata so reporting can be compared geographically.

All analysis happens inside rolling time windows. Every comparison on the site is made within the same window, so each story is evaluated against the reporting landscape at that specific moment.

Imbalance is therefore measured relative to current publishing activity, not long-term averages.

2. Story Clustering Across Languages

Reporting on the same event often looks different across languages and editorial styles.

Restworld uses cross-lingual semantic clustering to group related articles into a single story. This allows attention to be measured on one shared event rather than across fragmented language-specific copies.

3. Coverage Share and Imbalance

For each story inside the active window, Restworld measures regional share of attention.

In simple terms, the number of articles about a story from a given region is divided by that region’s total reporting output within the same window.

This means imbalance is not based on raw article counts. It is adjusted for overall publishing volume during that period.

Western share is then evaluated against the broader global distribution within that same window.

Stories must meet minimum coverage and regional diversity thresholds before imbalance labels are applied. This prevents sparsely covered or highly local events from being overstated.

Among eligible stories in each window, imbalance labels are assigned using a percentile ranking of attention ratios: stories in the bottom range are marked under-covered and stories in the top range are marked over-covered.

Imbalance is a relative measurement of attention at a particular point in time.

4. Rolling Updates

As new articles are ingested and the pipeline runs, the time window moves forward and metrics are recalculated.

Stories can shift between balanced, over-covered, and under-covered as reporting patterns change.

The system is designed to reflect live attention dynamics rather than static snapshots.

5. Limits

No global news dataset is complete.

Results depend on outlet inclusion, language mix, availability of public articles, and window definitions.

Restworld does not infer intent and does not judge editorial decisions. It measures patterns in attention based on available data.