Synthetic Clinical Dialogue Dataset for Ambient Scribing
收藏DataCite Commons2026-05-02 更新2026-05-07 收录
下载链接：
https://zenodo.org/doi/10.5281/zenodo.19986636
下载链接
链接失效反馈官方服务：
资源简介：
This dataset contains approximately 1,700 validated synthetic doctor-patient dialogue and structured clinical summary pairs, generated across nine batches as part of an MSc dissertation project on trustworthy small language models (SLMs) for ambient clinical scribing. Each sample comprises a multi-turn consultation dialogue and a corresponding SOAP-format clinical summary, produced by GPT-4o-mini (OpenAI) grounded with retrieval-augmented generation over curated medical guidelines.
The dataset is provided in three formats corresponding to the chat templates of the fine-tuned student models: Phi-3.5-mini-instruct, Llama-3.2-3B-Instruct, and Llama-3.2-1B-Instruct. Each format includes train, validation, and test splits.
No real patient data were used. All clinical scenarios were procedurally generated from randomised demographic and symptom parameters. The dataset is intended for research purposes in clinical NLP and knowledge distillation.
Associated code: https://github.com/Alirezaprogramerrd99/Ambient_Scribing_with_Synthetic_Data_Generator
提供机构：
Zenodo
创建时间：
2026-05-02
