2022 Volume 42 Issue 1 Pages 3-15
Clinical narratives contain important information such as symptoms and findings, and there is a need for technology to automatically extract this information. The development of practical technology requires a corpus with annotations that cover all the described information, while such a corpus does not exist at present. We have developed and published a corpus of case reports with comprehensive annotations on patient conditions. In this paper, we report on the establishment of the annotation criteria. As a result of repeatedly annotating case reports and modifying the annotation criteria, we obtained annotation criteria consisting of 50 entity types, 14 attributes, and 36 relations, which not only allow us to express more detailed information than previous studies, but also allow us to capture temporal changes in factuality that could not be expressed before.
