News Dashboard
Real-time trust-weighted news velocity across 93 international RSS sources.
What It Does
The News Dashboard is the fastest way to see where media attention is spiking globally. Countries appear as tiles sized by trust-weighted velocity z-score — so a large tile means an unusual surge in credible coverage, not just a high raw article count. A country with 3 Reuters articles can outscore a country with 30 low-trust articles.
Tiles glow red at critical levels (≥3σ) and amber at elevated activity (≥1.5σ). Click any tile to go deeper — past the article list, all the way down to the exact reasoning behind every country attribution in every article.
Data Source
News articles come from 93 international RSS feeds polled every 7 minutes by the News-Chef background worker. Each source carries a trust score from 0.3 to 1.0. Non-English headlines are machine-translated automatically. Each article is classified into a category (Military, Geopolitical, Health, Arts, General) and attributed to countries using NLP entity extraction before it enters the velocity calculation.
The z-score measures how unusual the current coverage level is relative to the historical baseline for that country — not raw volume. This means a consistently high-coverage country only scores high when it is receiving more than its own normal level of attention.
Reading the Grid
- Tile size: proportional to trust-weighted velocity z-score
- Red glow: z ≥ 3.0 — critical anomaly
- Amber glow: z ≥ 1.5 — elevated activity
- Small / no glow: near-baseline or below-baseline activity
- Category badges on tiles: the dominant category driving that country's score
- Live stats bar (bottom): total articles in window, active source count, time to next refresh
Controls
| Control | What it does |
|---|---|
| 1h / 6h / 24h | Changes the velocity window. 1h is most reactive to breaking events; 24h smooths short spikes. |
| Category pills (MILITARY / GEOPOLITICAL / HEALTH …) | Filters the tile grid and the drill panel simultaneously to one topic area. |
| Search bar | Filters the country grid by name. Also filters articles inside the open drill panel by keyword. |
Country Drill
Click any country tile to open the country drill. The left panel is a full intelligence summary of that country's current signal — not just article cards, but the analytical context around them:
- Velocity Signal: current articles/hr vs the baseline rate, percentage deviation, and z-score. France recently showed 22/hr vs a 1.7/hr baseline — +1,164% vs normal at z=7.67σ.
- Corroboration: green CORROBORATED badge when 3+ independent sources with trust ≥0.7 publish in the same window. NeoLexx applies a 1.2× velocity bonus at this threshold because independent corroboration is a stronger signal than volume alone.
- Trust Signal: breakdown of high / mid / low trust article counts with average trust score. MIXED TRUST vs HIGH TRUST tells you whether the spike is driven by credible sources or noise.
- Top Sources: ranked list of which outlets are contributing most articles in this window.
- Category Mix: colour bar showing the proportional split across categories.
The right panel shows the filterable article feed — category tabs with counts (ALL / MILITARY / GEOPOLITICAL / HEALTH / ARTS / GENERAL), a time window selector independent from the main grid, and article cards in a two-column layout. Cards marked OFFICIAL come from government or institutional sources.
Heuristic Analysis — The Glass Box
Click any article card to open the heuristic analysis panel. This is the deepest layer in NeoLexx — it shows you the exact reasoning the algorithm used to attribute that article to each country.
Color-coded entity extraction is applied directly to the headline and RSS summary. Every significant term is highlighted by signal type:
For the article "War in the Middle East: Israel says residential building hit by Iranian missile" — "Israel" and "Iranian" highlight blue as country entities, "War" highlights amber as a military category signal, and "missile" highlights as a named entity. You can see exactly what the algorithm read and why.
The Analysis Signals bar shows the count of Country, Category, and Entity signals extracted from the article.
The Country Weighting card lists every country attribution with full provenance:
- Country name and final attribution score
- Signal strength: STRONG (title match or multiple field matches) vs WEAK (summary-only mention)
- Which fields matched: title, summary, or both
- The exact terms that triggered the match — "iranian", "israel, israelis", "france"
An article from France 24 about an Iranian missile strike on an Israeli building will correctly attribute Iran (STRONG · title, matched "iranian") and Israel (STRONG · title + summary, matched "israel, israelis") at 0.80, while France scores lower (WEAK · summary, matched "france") because it only appears as the source outlet's name. This is the transparency layer — you are never asked to trust the attribution, you are shown the evidence.
Limits
- Coverage reflects what is in RSS headlines — not a direct measure of on-the-ground event intensity
- Trust scores are fixed at source level; an unusually strong article from a mid-trust outlet scores the same as a routine one
- The 7-minute refresh cycle means very breaking news may appear with a short lag
- NLP entity extraction can misattribute articles that mention a country incidentally — the heuristic panel lets you judge this yourself