Logo Lanfrica

Zekeriya-Ui/afyalugha

Domain:

natural language processinghealthcare

Record type:

softwaremodel
Creator:
Zek
Host:
Swahili-language AI chatbot for patient symptom triage and health information retrieval in coastal Kenya β€” Deploying African NLP for Healthcare Equity # πŸ₯ Afyalugha β€” Swahili AI for Patient Triage & Health Access > **Deploying Swahili Language AI for Patient Triage and Health Information Access in Coastal Kenya** **Afyalugha** (from Swahili: *afya* = health, *lugha* = language) deploys a Swahili-language AI chatbot for symptom triage and health information retrieval in coastal Kenya. The project addresses critical language barriers that prevent **60% of rural patients** from accessing accurate health information in English-only digital health tools. --- ## 🌍 The Problem Africa has over 2,000 languages, yet most AI systems are built on English corpora. Even Swahili β€” spoken by 200M+ people β€” lacks adequate NLP coverage for medical applications. In coastal Kenya: - **90%+ of coastal Kenyans** use Swahili as their primary language - **Almost all digital health tools** are English-only - Medical AI chatbots trained on Western corpora **misdiagnose 30% more frequently** in African languages - Current NLP models exhibit **44% lower translation accuracy** for Swahili medical text - Less than **1% of public Kenya-available NLP corpora** contain African medical text --- ## 🎯 Project Objectives | Objective | Target | |-----------|--------| | Medical NER accuracy | β‰₯85% (vs. 58% Masakani baseline) | | Unique patients reached | 50,000 in Year 1 | | Language barrier reduction | 40% reduction in self-reported barriers | | Triage accuracy improvement | β‰₯5% vs. English-only tools | | Open-source medical terms | 5,000+ terms under CC BY 4.0 | | Health facility deployments | 10 clinics (Mombasa, Kilifi, Kwale) | --- ## πŸ—οΈ System Architecture ``` β”Œβ”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β” β”‚ USER INTERFACES β”‚ β”‚ WhatsApp β”‚ SMS β”‚ Web App β”‚ Speech-to-Text β”‚ β””β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”¬β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”˜ β”‚ β”Œβ”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β–Όβ”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β” β”‚ SWAHILI NLP PIPELINE …