Multi-OphthaLingua
收藏资源简介:
Multi-OphthaLingua 是首个多语言眼科问答数据集,由全球具有多样化语言背景的眼科专家创建,涵盖了英语、西班牙语、菲律宾语、葡萄牙语、普通话、法语和印地语。该数据集包含1184个问题,涉及基础眼科科学、临床案例、手术实践和术后护理等多个主题。问题以多选题形式呈现,每个问题有四个选项,并经过母语为相应语言的眼科专家审查,确保了问题的中立性和文化适宜性。数据集的创建旨在评估和减少大型语言模型在低收入和中等收入国家(LMICs)中的语言偏差,促进全球范围内眼科问答任务的公平应用。
Multi-OphthaLingua is the first multilingual ophthalmic question-answering dataset, created by ophthalmologists with diverse linguistic backgrounds across the globe, covering English, Spanish, Tagalog, Portuguese, Mandarin, French, and Hindi. This dataset contains 1,184 questions spanning multiple topics including basic ophthalmic science, clinical cases, surgical practice, and post-operative care. The questions are presented in multiple-choice format with four options per question, and have been reviewed by ophthalmologists who are native speakers of the corresponding language to ensure the neutrality and cultural appropriateness of the questions. The dataset was developed to evaluate and mitigate linguistic biases of large language models (LLMs) in low- and middle-income countries (LMICs), and to promote the equitable application of ophthalmic question-answering tasks globally.

- 1Multi-OphthaLingua: A Multilingual Benchmark for Assessing and Debiasing LLM Ophthalmological QA in LMICs麻省理工学院 · 2024年



