Purpose/Significance To construct a Chinese medical scale corpus of core content elements, so as to provide a data basis for the task of extracting related knowledge entities and relations. Method/Process An annotation schema covering five types of scale-related entities such as scale names, measurement concepts, measurement items, and their corresponding codes, as well as four types of semantic relations is designed, and a unified annotation standard is formulated. A double-blind manual annotation approach is adopted to annotate semantically rich paragraphs from 1 491 Chinese core journal articles. The corpus quality is further evaluated through inter-annotator agreement and downstream task experiments, ultimately resulting in the CMedScale corpus. Result/Conclusion After introducing the corpus examples, the Micro-F1 scores of entity recognition in each model increased by 2.95 to 13.89 percentage points, and those for relation extraction improved by 16.93 to 33.33 percentage points. The CMedScale corpus provides high-quality data support for Chinese medical scale knowledge extraction and related downstream research tasks.