HoangHa/medical-data

Meddies Consultant 是一个高质量、结构化的医疗对话数据集，专注于越南语优先的临床咨询建模，同时包含英语数据以支持双语迁移。该数据集旨在解决现实世界多轮医患咨询中越南语医疗对话数据稀缺的问题。通过角色驱动的生成和临床框架（如FIFE和OPQRST）来多样化患者和咨询动态，以提高临床对话的覆盖范围、鲁棒性以及训练信号，确保在越南语语境下安全、结构化的咨询行为。数据集包含1,236种独特疾病，覆盖广泛的疾病范围，并采用临床结构化监督来提升对话质量。子集包括多轮咨询的英语和越南语对话，以及用于问答监督的RandomQA和RandomQuestion。Meddies Consultant is a high-quality, structured medical conversation dataset focused on Vietnamese-first clinical consultation modeling, while also including English data for bilingual transfer. It addresses the scarcity of realistic multi-turn doctor-patient consultation data in Vietnamese. By using persona-driven generation and clinical frameworks such as FIFE and OPQRST, the dataset diversifies patient and consultation dynamics to enhance clinical conversation coverage, robustness, and training signals for safe, structured consultation behavior in the Vietnamese context. The dataset covers a wide range of diseases with 1,236 unique target diseases and employs clinically structured supervision to improve dialogue quality. Subsets include multi-turn consultations in English and Vietnamese, as well as RandomQA and RandomQuestion for QA-style supervision.