CHIP2022-医疗文本诊疗决策树抽取任务

临床诊疗流程是医学核心知识，且在构建临床辅助诊疗系统、自动问诊机器人等应用中扮演着重要的作用。但是目前这类知识的结构化往往依赖于医学专家人工整理。本数据集定义的Text2DT任务旨在将临床诊疗流程建模为一个诊疗决策树，一个由条件节点和决策节点组成的树型结构，条件节点表示需要做出的条件判断，决策节点表示需要做出的诊疗决策。在这个任务中，研究人员不仅要进行实体与语义的抽取，也需要把多个信息表示为一个完整的决策流程。
数据集由华东师范大学语言认知与知识计算团队提供，并依托CHIP2022大会举办的“医疗文本诊疗决策树抽取任务”(Text2DT任务)。<br />
<br />
<font color="red">本次赛题数据已在CBLUE评测基准开放</font>：https://tianchi.aliyun.com/dataset/95414 ，欢迎研究人员开展诊疗决策树自动抽取方向的研究。

Clinical diagnosis and treatment workflow is core medical knowledge, and plays a pivotal role in applications such as clinical auxiliary diagnosis and treatment systems and automatic consultation robots. However, the structuring of such knowledge currently relies heavily on manual collation by medical experts.<br /><br />The Text2DT task defined by this dataset aims to model clinical diagnosis and treatment workflows into a diagnosis and treatment decision tree, a tree-structured framework composed of conditional nodes and decision nodes. Conditional nodes represent conditional judgments that need to be made, while decision nodes represent clinical diagnosis and treatment decisions that need to be made. In this task, researchers not only need to perform entity and semantic extraction, but also integrate multiple pieces of information to represent a complete decision-making workflow.<br /><br />This dataset is provided by the Language Cognition and Knowledge Computing Team of East China Normal University, and is based on the "Medical Text Diagnosis and Treatment Decision Tree Extraction Task (Text2DT Task)" held at the CHIP 2022 conference.<br /><br /><font color="red">The data for this competition task has been made publicly available on the CBLUE benchmark:</font> https://tianchi.aliyun.com/dataset/95414. Researchers are welcome to conduct research on the automatic extraction of diagnosis and treatment decision trees.

搜集汇总

背景与挑战

背景概述

CHIP2022-医疗文本诊疗决策树抽取任务数据集旨在从医疗文本中自动抽取诊疗决策树，包含500个标注样本（训练300，验证和测试各100）。任务要求同时进行实体关系抽取和决策流程构建，评估指标包括三元组F1分数、决策树准确率等。数据集由华东师范大学团队提供，适用于辅助诊疗系统和医疗教学研究。