kenya-agricultural-qa
收藏资源简介:
肯尼亚农业问答数据集是一个双语(斯瓦希里语和英语)数据集,旨在支持肯尼亚农业人工智能发展,解决AI模型在斯瓦希里语农业领域表现不佳的问题。数据内容涵盖作物病害诊断、市场价格信息、天气警报和农业最佳实践等农业咨询主题,适用于问答和文本生成任务。数据集规模较小,样本数少于1千,基于KALRO/KEPHIS指南的DEMO合成数据生成,由Gabriel Mahia创建,采用CC BY 4.0许可证发布。主要用于训练和评估面向肯尼亚800万小农户的AI农业咨询工具,是东非AI技术堆栈的一部分。
The Kenyan Agricultural Q&A Dataset is a bilingual dataset supporting the development of agricultural artificial intelligence in Kenya, comprising question-answer pairs in Swahili and English. This dataset aims to address the poor performance of AI models in the Swahili agricultural domain: studies have shown that AI models have an error rate four times higher in Swahili than in English, and specialized agricultural terminology in Swahili is nearly absent from the training data of standard large language models. The dataset covers agricultural advisory topics including crop disease diagnosis, market price information, weather alerts, and agricultural best practices, and is suitable for question-answering and text generation tasks. With fewer than 1,000 samples, it is generated from DEMO data based on the guidelines of KALRO/KEPHIS, created by Gabriel Mahia, and released under the CC BY 4.0 license. This dataset is primarily used to train and evaluate AI agricultural advisory tools for Kenya's 8 million smallholder farmers, and constitutes a part of the East African AI technology stack.
数据集概要
- 名称: Kenya Agricultural Q&A Dataset
- 许可证: CC BY 4.0
- 作者: Gabriel Mahia(主页:gabrielmahia.github.io)
- 数据来源: 基于KALRO/KEPHIS指南生成的DEMO合成数据
语言与规模
- 语言: 斯瓦希里语(sw)、英语(en)
- 规模: n<1K(少于1000条样本)
任务类型
- 问答(question-answering)
- 文本生成(text-generation)
领域标签
- 肯尼亚、农业、斯瓦希里语、东非、农耕、作物病害、粮食安全
内容覆盖
- 农作物病害诊断
- 市场价格
- 天气预警
- 农业最佳实践
研究背景
该数据集旨在支持斯瓦希里语农业人工智能的开发。研究表明,AI模型在斯瓦希里语中的错误率是英语的4倍(arXiv:2509.04516),且斯瓦希里语中的农业专业词汇在标准LLM训练数据中几乎为零。该问答对用于训练和评估面向肯尼亚800万小农户的AI咨询工具。
所属项目: East Africa AI stack(详见:gabrielmahia.github.io)




