MEDEC
收藏资源简介:
MEDEC是由微软和华盛顿大学联合创建的医疗错误检测与纠正基准数据集,旨在评估语言模型在临床文本中检测和纠正错误的能力。该数据集包含3848条临床文本,涵盖了诊断、管理、治疗、药物治疗和病原体等五类常见错误。数据来源于美国三家医院的临床记录,并通过两种方法生成错误:一种基于医学考试题目,另一种基于真实临床记录。数据集的应用领域主要集中在医疗文档的自动验证和错误纠正,旨在提高医疗文档的准确性和一致性,减少医疗错误对临床决策的影响。
MEDEC is a benchmark dataset for medical error detection and correction jointly created by Microsoft and the University of Washington, designed to evaluate the capability of language models to detect and correct errors in clinical texts. This dataset includes 3,848 clinical text entries, covering five common types of errors: diagnosis, management, treatment, pharmacotherapy, and pathogen-related errors. The data is sourced from clinical records of three hospitals in the United States, with errors generated via two approaches: one based on medical examination questions and the other based on real clinical records. The main application scenarios of this dataset focus on automatic verification and error correction of medical documents, aiming to improve the accuracy and consistency of medical documentation and reduce the impact of medical errors on clinical decision-making.

- 1MEDEC: A Benchmark for Medical Error Detection and Correction in Clinical Notes微软健康与生命科学人工智能, 华盛顿大学生物医学与健康信息学 · 2024年



