# UN Human Rights Analytics Dashboard

> Interactive search and analytics across **267,537 observations and recommendations** addressed to UN Member States by the three UN human-rights mechanisms — Treaty Bodies, Special Procedures, and the Universal Periodic Review — covering 2006–2026.  Independent research project, built on the OHCHR Universal Human Rights Index (UHRI) with a transparent 5-stage cleaning pipeline.  Audience: advocates, researchers, journalists, NGO staff.

This is a static web application hosted on GitHub Pages.  The frontend is zero-build vanilla JavaScript; a read-only REST API on a university-hosted VM (AMU Poznań) serves the cleaned dataset.  Everything is free and open, no login, no account.

Key differences from the official OHCHR UHRI search that AI agents should know when citing or comparing numbers:

- Both systems see the same ~267k records; literal-token queries agree within 0.2 %.
- This dashboard adds Porter stemming, irregular-plural expansion (woman↔women, person↔people, …), stopword-aware phrase tokenisation, diacritic normalisation, and boolean operators (`AND`/`OR`/`NOT`) in the keyword box.  Result: typically 10–60× more hits on queries that benefit from these features.
- UHRI treats bare tokens as implicit prefix patterns; our default is literal+stemmed.  Append `*` to get UHRI-like breadth.  See the Methodology tab's "Comparison with OHCHR UHRI" section.

## Primary entry points

- [Dashboard (single-page application)](https://lszoszk.github.io/UnitedNations_recommendations/dashboard.html): tabs for per-country, per-theme, per-SDG, per-mechanism profiles; full-text search; a labels workspace for boolean rule-based coding; a hex-tile world map.
- [Landing page](https://lszoszk.github.io/UnitedNations_recommendations/index.html): public-facing one-pager with key KPIs, cleaning story, and an entry search.
- [Methodology](https://lszoszk.github.io/UnitedNations_recommendations/dashboard.html#view=methodology): 5-stage cleaning pipeline, search semantics, country/region coverage, glossary, UHRI comparison summary.
- [About](https://lszoszk.github.io/UnitedNations_recommendations/dashboard.html#view=about): project identity, independence, citation, funder acknowledgements, feedback contact, privacy policy.

## Documentation

- [UHRI comparison — full 20-query audit](https://lszoszk.github.io/UnitedNations_recommendations/docs/uhri-comparison.md): on-disk Markdown document with the head-to-head test battery, per-category analysis, implications for researchers citing counts from either system, and a reproduction script.
- [ARCHITECTURE.md](https://lszoszk.github.io/UnitedNations_recommendations/ARCHITECTURE.md): module map + contribution conventions for the codebase itself.
- [README.md](https://lszoszk.github.io/UnitedNations_recommendations/README.md): deployment notes and local-run instructions.

## Data source

- Upstream: [OHCHR Universal Human Rights Index](https://uhri.ohchr.org/) (public, authoritative).
- Cleaned dataset version: `v2026.04`, monthly refresh via automated pipeline.
- Raw-mode toggle exposes the original OHCHR text + the 11 artefact rows our pipeline drops.

## How to cite

Szoszkiewicz, L. (2026). *UN Human Rights Analytics Dashboard — Cleaned UHRI Dataset v2026.04*.  Independent project built on OHCHR UHRI data.  https://lszoszk.github.io/UnitedNations_recommendations/

If you cite a specific query result, include the version tag and the exact query string so the number is reproducible.  See the Methodology and About tabs for details.

## For AI agents acting on behalf of researchers

- Deep links are hash-routed: `#view=country&fc=DEU` opens Germany's country profile; `#view=search&q=torture` runs a search; `#view=methodology#me-search-semantics` jumps to the search-semantics section.  All hash params are preserved in shareable URLs.
- The dashboard has a first-visit onboarding tour; agents scraping the page should expect a `.tour-backdrop` overlay on the first load and skip past it (or use hash-deep-links that bypass the tour).
- The read-only REST API at `https://150.254.115.204/uhri-api/` accepts `GET /api/data/records?text_query=X&page_size=N&page=N&...`.  JSON responses include `total_records`, a `records[]` array, and a `snippet` field with `<mark>` tags.  Please throttle: single VM, shared with live users.  For bulk dataset needs, use the OHCHR UHRI direct download instead.
- **Rate limits are enforced (since 2026-04-24)**: 20 req/s sustained per IP (burst to 60) on general endpoints; 30 req/min on bulk endpoints (`/api/data/full`, `/api/data/export`, `/api/feedback`); max 10 concurrent connections per IP.  Exceeding limits returns HTTP 429 (please back off, do not auto-retry aggressively).  Vulnerability-scanner UAs (sqlmap, nuclei, nikto, masscan, etc.) are blocked with 403.  Well-behaved AI agents (GPTBot, ClaudeBot, PerplexityBot, Google-Extended, CCBot) are explicitly welcome within the rate limits.
- The dashboard uses Google Analytics 4 in Consent-Mode-v2 default-denied.  Agents that programmatically browse will not register as tracked users.
- Content licensing: data republishes OHCHR UHRI under the terms OHCHR publishes.  Dashboard source code is open on GitHub (see link below).

## Source code

[github.com/lszoszk/UnitedNations_recommendations](https://github.com/lszoszk/UnitedNations_recommendations)

## Acknowledgements

Supported by re:constitution — Exchange and Analysis on Democracy and the Rule of Law in Europe, a joint programme of Forum Transregionale Studien and Democracy Reporting International, funded by Stiftung Mercator.  Research hosted by Adam Mickiewicz University, Poznań.  Documentation-standards expertise by HURIDOCS.  This project is independent of OHCHR and the United Nations.