Evolution of semantic annotations (ELISA)

在医学领域使用知识组织系统（KOS），如叙词表或本体，在解决语义互操作性问题方面显示出巨大的价值。在许多情况下，KOS用于注释现实世界的对象，如电子健康记录（EHR）、临床试验的病例报告表格（CRF）、基因或出版物，以便使其语义对应用程序和工具明确。医学知识的动态性迫使知识工程师不断修改kss或基础数据的内容。这些更改可能直接影响与更改的概念或更改的数据相关联的注释。因此，新的KOS版本可以使以前的注释无效。类似地，如果修改了注释文档的内容，则现有注释可能变得无关紧要。因此，许多注释可能会失去其相关性和价值，从而阻碍了对注释数据的预期使用和利用。因此，ELISA（语义注释的进化）项目将开发方法和工具来支持（半）自动维护受KOS或信息进化影响的语义注释，以保持注释的可利用性。所提出的方法将基于一个全面的注释模型，并利用来自KOS的细粒度信息以及信息演化来定义一个健壮的维护机制，以保持语义注释随时间的有效性。提议的框架将包括不同的方法来进化或迁移受更改影响的注释，可以直接修改注释，也可以在不能直接更改注释时间接修改，例如，在加密的电子病历中。ELISA方法将使用文档语料库（如ehr或CRFs）和大型生物医学KOS（如SNOMED CT、NCI词库和ICD-10）之间的注释进行经验评估。专家验证的注释将从卢森堡电子研究平台以及莱比锡文明疾病研究中心（LIFE）的实际案例研究中获得。

The use of Knowledge Organization Systems (KOS) such as thesauri or ontologies in the medical field has shown great value to tackle semantic interoperability issues. In many cases, KOS are used to annotate real world objects such as electronic health records (EHR), case report forms (CRF) of clinical trials, genes or publications in order to make their semantics explicit for applications and tools. The dynamic nature of medical knowledge forces knowledge engineers to continuously revise the content of either KOSs or underlying data. These changes may directly impact the annotations associated with changed concepts or changed data. Hence, new KOS versions can invalidate previous annotations. Similarly, if the content of an annotated document is modified the existing annotation may become irrelevant. As a result, many annotations can lose their relevance and value thus impeding the intended use and exploitation of annotated data. The ELISA (Evolution of Semantic Annotations) project therefore will develop methods and tools for supporting the (semi-)automatic maintenance of semantic annotations affected by KOS or information evolution in order to keep annotations exploitable. The proposed approaches will be based on a comprehensive annotation model and make use of fine-grained information coming from KOS as well as information evolution to define a robust maintenance mechanism for preserving the validity of semantic annotations over time. The proposed framework will include different approaches to evolve or migrate annotations affected by changes, either by modifying annotations directly or an indirect way when annotations cannot directly be changed, e.g., in the case of encrypted EHRs. The ELISA approaches will be empirically evaluated using annotations between corpora of documents such as EHRs or CRFs and large biomedical KOS like SNOMED CT, NCI thesaurus and ICD-10. Expert-validated annotations will be available from real case studies in the Luxembourgish eSanté platform as well as the Leipzig Research Centre for Civilization Diseases (LIFE).