BanglaMedQA and BanglaMMedBench
收藏资源简介:
BanglaMedQA和BanglaMMedBench是首个大规模的孟加拉语生物医学多项选择题(MCQ)数据集,旨在评估医疗人工智能(AI)中的推理和检索能力。这两个数据集总共包含2000个问题,BanglaMedQA包含1000个来自孟加拉国医学院入学考试的真实问题,每个问题都附有正确答案的推理。BanglaMMedBench则包括另外1000个情境和复杂问题,这些问题是通过对英语MMedBench数据集进行翻译和精炼得到的。数据集的创建过程包括从孟加拉国医学院入学考试中收集问题,并进行筛选以确保质量和一致性。BanglaMMedBench的创建则涉及将英语数据集翻译成孟加拉语,并进行必要的预处理。这些数据集的应用领域是评估检索增强生成(RAG)策略的有效性,并为孟加拉语医疗问答系统的发展奠定基础。
BanglaMedQA and BanglaMMedBench are the first large-scale Bengali biomedical multiple-choice question (MCQ) datasets designed to evaluate reasoning and retrieval capabilities in medical artificial intelligence (AI). Collectively, these two datasets contain 2,000 questions in total. BanglaMedQA includes 1,000 real questions from Bangladesh's medical college admission tests, each accompanied by the reasoning for the correct answer. BanglaMMedBench, on the other hand, encompasses another 1,000 contextual and complex questions that were developed by translating and refining the English MMedBench dataset. The creation of BanglaMedQA involved collecting questions from Bangladesh's medical college admission tests and filtering them to ensure quality and consistency. For BanglaMMedBench, the creation process included translating the English dataset into Bengali and performing necessary preprocessing. These datasets can be used to evaluate the effectiveness of retrieval-augmented generation (RAG) strategies and lay a solid foundation for the development of Bengali medical question answering systems.
- 1BanglaMedQA and BanglaMMedBench: Evaluating Retrieval-Augmented Generation Strategies for Bangla Biomedical Question Answering伊斯兰技术大学计算机科学与工程系 · 2025年



