Cornet Ronald, Van Eldik Armand, De Keizer Nicolette
Dept of Medical Informatics, Academic Medical Center, Amsterdam, The Netherlands.
Stud Health Technol Inform. 2012;180:245-9.
Automated encoding of free-text clinical narratives using concepts from terminological systems is widely performed. However, the majority of natural language processing (NLP) tools and terminological systems involve the English language. As parts of the NLP process are language independent, and tools for various languages are available, an overview is needed to determine the applicability to performing NLP of Dutch medical texts. To this end an inventory of tools is created. A literature study and internet search were performed to describe available components for a Dutch NLP system, enabling to encode Dutch text as structured SNOMED CT output without the need to translate SNOMED CT in Dutch. We have found 31 papers, describing a variety of NLP frameworks and tools for the various NLP components for processing English and Dutch free text. Most of them are suitable for English free text, some of them are (also) usable for Dutch. To enable automated encoding of Dutch free text narratives, further research is needed to create a spelling checker, a negation detector, a domain-specific abbreviation/acronym list, and a concept mapper (to map Dutch terms to concepts in a terminological system). Furthermore evaluation of performance for the Dutch 'medical' language is needed.
使用术语系统中的概念对自由文本临床叙述进行自动编码的操作已广泛开展。然而，大多数自然语言处理（NLP）工具和术语系统都涉及英语。由于NLP过程的某些部分与语言无关，并且有适用于各种语言的工具，因此需要进行概述以确定其对荷兰医学文本执行NLP的适用性。为此创建了一个工具清单。通过文献研究和互联网搜索来描述荷兰语NLP系统的可用组件，从而能够将荷兰语文本编码为结构化的SNOMED CT输出，而无需将SNOMED CT翻译成荷兰语。我们找到了31篇论文，它们描述了用于处理英语和荷兰语自由文本的各种NLP组件的多种NLP框架和工具。其中大多数适用于英语自由文本，有些（也）可用于荷兰语。为了实现荷兰语自由文本叙述的自动编码，需要进一步研究以创建一个拼写检查器、一个否定检测器、一个特定领域的缩写/首字母缩略词列表以及一个概念映射器（将荷兰语术语映射到术语系统中的概念）。此外，还需要对荷兰语“医学”语言的性能进行评估。
