Skip to main content
. 2022 Sep 23;13(39):11487–11495. doi: 10.1039/d2sc04322j

F1-score of abbreviation detection, CNER, and POS tagging for six BERT-based models, including BatteryBERT, BatteryOnlyBERT, base BERT, as well as that for ChemDataExtractor v2.0 and v2.1.

Model Abbreviation detection CNER POS tagging
BatteryBERT-cased 0.9502 0.9584 0.9667
BatteryBERT-uncased 0.9475 0.9578 0.9632
BatteryOnlyBERT-cased 0.9516 0.9589 0.9640
BatteryOnlyBERT-uncased 0.9492 0.9598 0.9605
BERT-base-cased 0.9491 0.9458 0.9669
BERT-base-uncased 0.9464 0.9517 0.9633
ChemDataExtractor v2.0 0.6998 0.8649
ChemDataExtractor v2.1 0.8882 0.8649