SEMANTIC PARSER FOR MEDICAL FREE TEXT

描述(摘自申请摘要)：综合临床数据 基地是临床信息系统计算机化的基石。AS 这些系统不断发展，以支持越来越多的医疗保健服务 在这个国家，这些数据库的完整性和准确性将 重要性的提高。捕获临床数据的需求也将如此 它们以计算机可以分析和处理的形式存储，编码 形式。目前捕获的许多信息都不符合这一标准；它 是以各种自由文本医疗文档的形式收集的。这很自然 语言数据只能由人类读者来解释。 这项提案描述了旨在开发一种自然语言的研究 理解专门针对提取相关临床信息的系统 来自医学自由文本的事实。它建立在两个领域之前的工作的基础上。 在放射学领域已经开发了一种应用程序，该应用程序编码 相关临床数据采用句子语义模型。这 该项目寻求通过开发更多的上下文模型来扩展这项工作 其焦点是整个X射线报告的语义。 在诊断编码领域，以前的工作已经演示了一个 对承认诊断的自由文本进行编码的有前景的方法 语义模型来源于放射学中的工作。这一项目旨在 通过开发技术来扩展这项工作，以1)管理对常规 培训以更新本系统中的知识结构，以及2)扩展 帮助识别拼写错误和变体的语义模型 接受的缩写。 作为该项目的一部分，将开发两个系统来探索这些 自然语言理解问题的两个互补部分。 这些系统将接受一系列测试程序，作为 形成性评价。最终，该项目的目标是进一步 允许将捕获的医疗信息编码为 将自由文本转换为适合研究、质量保证、 医疗企业的管理和直接的临床决策支持。

DESCRIPTION (Taken from application abstract): Integrated clinical data bases are the cornerstones of computerized clinical information systems. As these systems evolve to support more and more of the health care delivered in this country, the completeness and accuracy of these data bases will increase in importance. So too will the need to capture the clinical data that they store in a form that computers can analyze and manipulate, a coded form. Much of the information currently captured fails this criterion; it is collected as a variety of free-text medical documents. This natural language data can be interpreted only by human readers. This proposal describes research intended to develop a natural language understanding system specifically aimed at extracting relevant clinical facts from medical free-text. It builds on previous work in two domains. In the domain of radiology an application has been developed that encodes relevant clinical data using a model of the semantics of sentences. This project seeks to extend that work by developing additional contextual models whose focus is the semantics of the entire x-ray report. In the domain of diagnostic coding, previous work has demonstrated a promising approach to encoding free-text admitting diagnoses using a semantic model derived from the work in radiology. This project seeks to extend that work by developing techniques to 1) manage the need for regular training to update knowledge structures in this system and 2) extend the semantic model to assist in recognizing misspellings and variations on accepted abbreviations. As a part of this project, two systems will be developed to explore these two complimentary parts of the natural language understanding problem. These systems will undergo a sequence of testing procedures as a part of formative evaluations. Ultimately, the goal of this project is to further techniques that allow the encoding of medical information captured as free-text into a form appropriate for research, quality assurance, the management of medical enterprises, and direct clinical decision support.