可概括的临床笔记部分识别与大型语言模型
Weipeng Zhou1, Timothy A Miller2,3
1Department of Biomedical Informatics and Medical Education, School of Medicine, University of Washington-Seattle, Seattle, WA 98195, United States.
JAMIA open
|August 14, 2024
概括
大型语言模型 (LLM) 显示出临床笔记部分识别的前景,GPT-4实现了高精度. 使用特定示例进行微调进一步提高了性能,使LLM几乎为此任务做好了生产准备.
科学领域:
- 自然语言处理自然语言处理.
- 临床信息学 临床信息学
- 医疗保健中的人工智能
背景情况:
- 临床笔记部分的识别对于信息检索和下游NLP任务至关重要.
- 传统的监督方法在不同临床数据集的可转移性方面面临挑战.
- 大型语言模型 (LLM) 为克服这些局限性提供了一个潜在的解决方案.
研究的目的:
- 评估LLM在临床注释部分识别方面的有效性.
- 为了比较各种LLM的性能,包括GPT-4,GPT-3.5和开源模型.
- 调查微调数据集大小和特异性对LLM绩效的影响.
主要方法:
- 使用自由文本部分定义,将框架部分识别作为一个问题答案任务.
- 在没有事先培训的情况下,评估了多个现成的LLM.
- 使用不同大小和特异性的数据集,微调精选的LLM.
主要成果:
- GPT-4获得了最高的F1得分 (0.77),超过了其他车型.
- 对于特定的切口类型,GPT-4显示出高精度 (F1>0.9为33%,F1>0.8为56%).
- 微调模型显示,较大的一般数据集的回报率下降,但在特定的部分识别示例中得到改善.
结论:
- LLM,特别是GPT-4,对可泛化临床笔记部分识别非常有希望,并且正在接近生产准备.
- 开源LLM正在迅速改善,并接近领先的专有模型的性能.
- 通过将部分识别示例纳入LLM微调数据集,可以实现进一步的改进.
相关概念视频
Mechanistic Models: Overview of Compartment Models
Mechanistic models, a category encompassing both physiological and compartmental modeling, differ from empirical models' approaches to incorporating known factors about the systems being modeled. Empirical models describe data with minimal assumptions, while mechanistic models aim to provide a robust description of available data by specifying assumptions and integrating known factors about the system. Compartmental analysis is a key example of a mechanistic model in pharmacokinetics and...
Concepts and Prototypes
The human nervous system handles vast amounts of information by translating sensory stimuli into neural impulses, which the brain processes, creating thoughts expressed through language or stored as memories. The brain also synthesizes information from emotions and memories, which significantly influence thoughts and behaviors. This intricate process creates a comprehensive mental picture.
The brain organizes this information using concepts, which are mental categories grouping linguistic data,...
The brain organizes this information using concepts, which are mental categories grouping linguistic data,...


