electricsheepafrica/chewie-community-health-worker-instruct

Chewie Instruct是一个双语（英语-斯瓦希里语）医疗指令数据集，旨在训练大型语言模型（LLMs）作为非洲社区健康工作者（CHWs）的助手。数据集包含约3,000多个高质量指令，涵盖初级医疗、分诊、母婴健康、传染病（疟疾、HIV/TB）、非传染性疾病和紧急危险信号等多个方面。数据来源基于WHO/卫生部临床指南的合成生成，经过AI整理和协议遵守性审查。数据集结构包括每个示例的指令和输出，遵循评估->行动->建议的协议。数据集的目的是微调模型以遵守CHW协议、检测危险信号、弥合语言差距并显示同理心。

Chewie Instruct is a bilingual (English-Swahili) medical instruction dataset designed to train large language models (LLMs) to act as assistants for Community Health Workers (CHWs) in Africa. The dataset contains approximately 3,000+ high-quality instructions covering primary healthcare, triage, maternal & child health, infectious diseases (Malaria, HIV/TB), NCDs, and emergency danger signs. The data is synthetically generated based on WHO/Ministry of Health Clinical Guidelines, curated by AI and reviewed for protocol adherence. Each example in the dataset includes an instruction and output, following the Assessment -> Action -> Advice protocol. The primary goal is to fine-tune models to adhere to CHW protocols, detect danger signs, bridge the language gap, and display empathy.