Skip to main content
. 2022 May 10;19(5):2584–2595. doi: 10.1109/TCBB.2022.3173562

TABLE 6. Performance Evaluation on an Independent Sample of 3000 Articles in litcovid.

Label-based measures Instance-based measures Time measures in the production environment (in sec)
Macro-F1 Macro-AP Micro-F1 Micro-AP F1 Accuracy
Binary BERT 0.8116 0.7251 0.8570 0.7733 0.8719 0.6327 8912.22
(2.97 sec/doc)
Linear BERT 0.8299 0.7449 0.8539 0.7701 0.8759 0.6354 1536.24
(0.51 sec/doc)
LitMC-BERT (ours) 0.8413 0.7690 0.8628 0.7813 0.8862 0.6512 1673.98
(0.56 sec/doc)

Time was measured in the production environment where only cpus are available.