Extended Methods and Software Development for Health NLP

健康 NLP 的扩展方法和软件开发

• 批准号: 10413157
• 负责人: Steven Bethard
• 金额: $ 44.54万
• 依托单位: BOSTON CHILDREN'S HOSPITAL
• 依托单位国家: 美国
• 财政年份: 2016
• 资助国家: 美国
• 起止时间: 2016-01-01 至 2025-05-31
• 项目状态: 未结题
• 来源: https://reporter.nih.gov/project-details/10413157
• 关键词: Address; Adult; Algorithms; Apache; Area; Attention; Big Data; Childhood; Clinical; Communities; Community Health; Computer software; Computerized Medical Record; Data; Data Set; Detection; Development; Environment; Evaluation; Foundations; Funding; Goals; Health; Healthcare; Information Retrieval; Institution; Investigation; Joints; Knowledge; Knowledge Extraction; Machine Learning; Manuals; Medical; Medicine; Methodology; Methods; Modeling; Modernization; Names; Natural Language Processing; Neural Network Simulation; Ontology; Patients; Penetration; Personal Satisfaction; Phenotype; Physicians; Public Health; Publications; Research; Risk; Solid; Source; Supervision; System; Techniques; Text; Time; Training; Training and Infrastructure; Uncertainty; Unified Medical Language System; Vision; Work; commercialization; data streams; deep learning; deep learning algorithm; design; electronic data; federated learning; health data; health knowledge; insight; learning strategy; medical specialties; method development; multitask; neural network; novel; online community; open source; point of care; precision medicine; programs; social media; software development; structured data; tool; translational medicine

Project Summary Our program vision is to unravel the information buried in health-related narratives by advancing text-processing methods in a unified way across all the genres of health texts and distributing them through an advanced NLP software platform under solid governance and sustainability. The crosscutting theme is the investigation of methods for health NLP made possible by big data, fused with health knowledge. The underlying theme of this renewal is the development of methods towards generalizable, efficient and knowledge-rich models in the context of modern machine learning techniques, particularly models implementing attention mechanisms and using large unlabeled datasets. There is growing penetration of deep learning approaches in the field of health natural language processing. Our proposal aims to address critical methodological gaps and understudied areas in the current unprecedented fast-paced environment. Therefore, our renewal lays out novel and much needed explorations of health NLP research which we will advance through our specific aims. Our datasets will continue to span the spectrum of health-related data – Electronic Medical Records clinical narrative, patient-authored on- line community posts, and health-related social media. The evaluation of the methods we will develop will be performed on the key clinical tasks of concept extraction, relation extraction, and phenotyping with comparisons to other traditional or deep learning algorithms as baselines. We will demonstrate impact of our methods and tools through several use cases, ranging from clinical point of care to public health, to translational and precision medicine. Finally, we will disseminate our work through community activities to advance the state of the art in health natural language processing.

项目摘要 我们的计划愿景是通过推进文本处理来解开隐藏在健康相关叙述中的信息 方法以统一的方式跨越所有体裁的健康文本，并通过高级NLP分发它们 稳固治理和可持续发展下的软件平台。贯穿始终的主题是对 大数据与健康知识融合的健康NLP方法。这件事的基本主题 更新是向可概括的、高效的和知识丰富的模型的方法的发展 现代机器学习技术的背景，特别是实现注意机制和 使用大量未标记的数据集。深度学习方法在卫生领域的渗透率越来越高 自然语言处理。我们的建议旨在解决关键的方法差距和研究不足的领域。 在当前前所未有的快节奏环境下。因此，我们的更新布局新颖而又亟需 我们将通过我们的具体目标来推进健康NLP研究的探索。我们的数据集将继续 要跨越与健康相关的数据范围-电子病历临床叙述、患者创作的- LINE社区帖子和与健康相关的社交媒体。我们将开发的方法的评估将是 完成了关键的临床任务，如概念提取、关系提取和表型比较 以其他传统或深度学习算法为基线。我们将展示我们的方法和 工具贯穿多个使用案例，从临床护理到公共卫生，再到翻译和精确度 医药。最后，我们将通过社区活动宣传我们的工作，以促进 健康自然语言处理。