CHIP2023

CHIP2023-中文医学小样本命名实体识别 中文医学文本小样本命名实体识别数据集(Chinese Medical text Few-Shot Entity Recognition Dataset，CMFD)包含15种标签：item、sociology、disease、etiology、body、age、adjuvant therapy、electroencephalogram、equipment、drug、procedure、treatment、microorganism、department、epidemiology、symptom和others(不属于实体类型的一种)。数据集考虑了实体类型长尾分布的问题。小样本学习是机器学习的一种特殊情况，限制了目标任务的训练数据量。对于N-way-K-shot命名实体识别任务（N=5，K=1或5），每条数据包含N个类型，且每个类型至少K个实例，构成该数据的support set用于训练，此外还包括相应的query set用于测试。CHIP2023-Chinese Medical Few-Shot Named Entity Recognition The Chinese Medical Text Few-Shot Entity Recognition Dataset (CMFD) includes 15 entity labels: item, sociology, disease, etiology, body, age, adjuvant therapy, electroencephalogram, equipment, drug, procedure, treatment, microorganism, department, epidemiology, symptom, and others (a category for instances that do not belong to any predefined entity type). The dataset addresses the long-tailed distribution problem of entity types. Few-shot learning is a specialized paradigm within machine learning that limits the size of the training dataset for target tasks. For the N-way-K-shot named entity recognition task (with N=5 and K=1 or 5), each task episode comprises N entity types, with a minimum of K instances per type, which constitutes the support set for model training, alongside a corresponding query set for model evaluation.

搜集汇总

背景与挑战

背景概述

CHIP2023是一个中文医学文本小样本命名实体识别数据集，包含15种标签，适用于N-way-K-shot命名实体识别任务，支持小样本学习场景。