Logo Lanfrica

SMA75237/lingua-patria

Domaine:

natural language processing

Type de record:

dataset
Créateur:
SMA
Hôte:
Multilingual parallel corpus on African constitutional governance & anti-corruption · EN/FR/PT · DPG Nominee · GOVBENCH AFRICA # LINGUA PATRIA — Open Multilingual NLP Corpus (African Governance) --- ## Overview **LINGUA PATRIA** is a sovereign, African-produced open multilingual corpus for NLP research and evaluation, focused on African governance, public finance, and institutional accountability. The **Open Subset v1.0** — the Digital Public Good artefact — comprises **77 parallel segments** (≈ 21,000 words per language) across **English, French, and Portuguese**, drawn from **19 sources** : 6 peer-reviewed working papers published in three language versions (EN/FR/PT), and the monograph *La Révolution patrimoniale — Biens publics & Troisième indépendance africaine* (Armand Salouo, Kurukan Publishing, 2026). The corpus underpins **GOVBENCH AFRICA**, a benchmark for evaluating language model performance on African governance tasks. --- ## Repository Contents | File | Description | |---|---| | `lingua_patria_open_subset.csv` | Open Subset v1.0 — 77 segments EN/FR/PT, CC BY 4.0 | | `DATACARD.md` | Full dataset card (HuggingFace format, YAML frontmatter) | | `RESPONSIBLE_USE_POLICY.md` | Do No Harm policy — 10 sections, co-signed Kurukan + Transparency Africa | | `RESPONSIBLE_USE_POLICY.pdf` | Signed PDF version | --- ## SDG Alignment | Goal | Relevance | |---|---| | **SDG 16** — Peace, Justice & Strong Institutions | Corpus centred on public governance, accountability, and asset declaration systems | | **SDG 10** — Reduced Inequalities | Open multilingual infrastructure for underrepresented African languages and policy contexts | | **SDG 4** — Quality Education | Open academic resource for African NLP research communities | --- ## FAIR Principles | Principle | Implementation | |---|---| | **Findable** | DOI: `10.5281/zenodo.20578745` · HuggingFace: `kurukanpublishing/lingua-patria` · GitHub: `kurukan/lingua-patria` | | **Accessible** | CC BY 4.0 — no registration required · CSV format · Human and machine readable | | **Interoperable** | Croissant metadata · YAML data card · …