Logo Lanfrica
  • Accueil
  • Atlas
  • Analyses
  • Documentation
  • Sign in

© 2026 Lanfrica. Tous droits réservés. Tous les droits d'auteur des ressources affichées sur le site Web Lanfrica appartiennent aux détenteurs de droits d'auteur d'origine, sauf indication contraire explicite.

drjforrest/african-ai-strategies

Domaine:

natural language processingdigital infrastructure

Type de record:

project
Créateur:
drj
Hôte:
An analysis of national governance strategies for AI in African countries # African AI strategies × WHO digital-health competencies Quantitatively maps African national **AI strategies** onto the competency domains of the **WHO Global Strategy on Digital Health** (2020–2027), using multilingual sentence-transformer embeddings, and visualises the result as heatmaps. For each country, the strategy text is chunked into passages, every passage is embedded and compared (cosine similarity) against each WHO competency-domain description, and the passage scores are aggregated into one score per (country, competency). The output is a country × competency alignment matrix. > **What the score means.** It measures how strongly a document *engages with the > WHO digital-health competencies in text* — not a country's real-world > digital-health capability, and not AI-strategy quality in general. A horizontal > AI policy that barely mentions health will score low even for a country that > leads in digital health in practice (see *Two analysis tracks* below). --- ## Two analysis tracks The corpus contains two genres of document, analysed separately so they stay comparable: | Track | Corpus | What it answers | |---|---|---| | **Primary** | National **AI strategies** (`data/national_strategies/`) | How much does each country's AI strategy engage digital health? | | **Secondary** | Dedicated **digital-health / eHealth strategies** (`data/digital_health_strategies/`) | How well do purpose-built digital-health strategies align with the WHO framework? | A digital-health document aligns with WHO *digital-health* competencies almost by construction, so mixing the two genres breaks comparability. Keeping them separate is what makes each track interpretable. **Rwanda is the canonical example:** its horizontal *National AI Policy* ranks last in the primary track (0.43 — it barely mentions health), while its dedicated *Digital Health Strategy* scores 0.61 in the secondary track. Same country, two documents, two correct answers. --- ## Quickstart ```bash # …

Visit

github.com

Tasks

information extraction