Knowledge Map

Real-time global intelligence dashboard. Aggregates live signals from Reddit and Wikipedia, maps them to countries, and renders the result as an interactive choropleth — entirely client-side, zero infrastructure.

Overview

Knowledge Map fetches trending content from five Reddit feeds and Wikipedia's most-read articles, performs entity recognition against a curated dictionary of ~120 countries (including aliases and demonyms), scores each country by weighted mention volume, and paints a Leaflet-powered world map in real time.

No backend. No API keys. No build step. Three files, open in a browser, done.

Quick Start

# Option A — any static file server
npx -y http-server . -p 8080

# Option B — Python
python3 -m http.server 8080

Open http://localhost:8080. Data loads automatically on first paint.

Note: Opening index.html directly via file:// will fail in most browsers due to CORS restrictions on the fetch calls to Reddit and Wikipedia.

Architecture

index.html    Semantic markup, ARIA roles, responsive meta
style.css     Design tokens (CSS custom properties), layout, responsive breakpoints
script.js     All runtime logic — fetch, NLP pipeline, Leaflet bindings, UI state

No frameworks. No transpilation. No package.json.

Data Pipeline

Reddit JSON API ──┐
                   ├──▶ normalize scores ──▶ detect countries ──▶ extract topics ──▶ classify category
Wikipedia REST ───┘                              │                      │                    │
                                                 ▼                      ▼                    ▼
                                          country scores          TF-ICF ranking        filter by tab
                                                 │
                                                 ▼
                                        z-score normalization ──▶ intensity ──▶ color interpolation ──▶ render

Entity recognition uses a pre-built lookup map of country names, ISO aliases, capital cities, and demonyms — sorted by key length (longest-first) to avoid partial matches. A deny list filters out false positives like "Swiss cheese", "French toast", and "Korean BBQ".

Scoring applies log10 normalization to Reddit upvotes and median-relative scaling to Wikipedia views before combining them. The final per-country intensity is derived from z-score normalization across all active countries, mapped to [0, 1].

Topic ranking uses TF-ICF (Term Frequency × Inverse Country Frequency) — a variant of TF-IDF adapted for the country dimension instead of documents. Topics that appear in many countries are down-weighted; topics concentrated in a single region are surfaced.

Features

Feature	Description
Live choropleth	Animated color transitions via requestAnimationFrame lerping between hex values
Detail panel	Score, trending topics (bar chart), sources with metadata, topical neighbours via cosine similarity
Category filters	Conflict & Politics · Science & Tech · Economy · Culture & Sports
Trending ticker	Top 20 global topics, click to flash-highlight related countries
Search	Fuzzy country lookup sorted by current activity score, keyboard-navigable
Freshness indicator	Green / amber / red dot reflecting data age
Mobile drawer	Bottom-sheet with drag-to-dismiss on touch devices
Keyboard shortcuts	/ search · R refresh · F toggle filters · Esc close panel

Data Sources

Source	Endpoint	Auth	Rate Limits
Reddit	/r/{subreddit}/hot.json	None	~60 req/min per IP
Wikipedia	Wikimedia REST /metrics/pageviews/top	None	Generous for single-user

Subreddits polled: worldnews (100), popular (50), news (50), technology (50), science (50).

Data refreshes every 5 minutes (SYNC_INTERVAL). Each user's browser fetches independently — there is no shared state.

Limitations

• English-language bias. The pipeline only processes English-language Reddit and Wikipedia. Events not covered in English media will not appear.
• No contextual disambiguation. "French cuisine" registers as France activity. The deny list mitigates common false positives but is not exhaustive.
• Client-side only. No persistence across sessions, no shared view between users, no historical data beyond the current session.
• Wikipedia daily lag. The mostread endpoint reflects views accumulated since midnight UTC. Results improve throughout the day.
• Reddit rate limits. During high-traffic periods, Reddit's public API may return 429s. The app degrades gracefully — it continues with whatever sources responded and notifies the user.

Browser Support

Requires: fetch, async/await, Promise.allSettled, CSS custom properties, backdrop-filter, Clipboard API.

Tested: Chrome, Firefox, Safari, Edge (current stable). No IE support.

External Dependencies

All loaded from CDN at runtime — no install step.

Library	Version	CDN
Leaflet	1.9.4	unpkg
Font Awesome	6.5.1	cdnjs
Inter (font)	Variable	Google Fonts
GeoJSON boundaries	—	johan/world.geo.json

Configuration

Constants at the top of script.js:

const SYNC_INTERVAL = 5 * 60 * 1000;  // refresh cadence (ms)
const GEOJSON_URL   = '...';           // country boundaries source

Lower SYNC_INTERVAL for faster updates. Stay above 60s to avoid Reddit rate limits.

License

MIT