Skip to main content
. 2020 May 4;8(5):e17637. doi: 10.2196/17637

Table 2.

Components of the two datasets.

Dataset Number of records per set

Total Training set Validation set Test set
CEMRa dataset 4000 2400 800 800
CCKSb 2018 1000 600 N/Ac 400

aCEMR: Chinese electronic medical record.

bCCKS: China Conference on Knowledge Graph and Semantic Computing.

cNot applicable; because the comparison method does not divide the validation set on the CCKS dataset, we have kept this the same as the original experiment to make the comparison fair.