Fig. 3
From: Chinese medical named entity recognition integrating adversarial training and feature enhancement
(a) BERT feature extraction process: step.1 Segmentation of sentences; step.2 Full word mask; step.3 Add special tokens, [CLS] for sentence beginnings, [SEP] for separating sentences; step.4 Output embedding; step.5 Transformer feature extraction. (b). Embedded representation of BERT10.
