Research Article

Medical Specialty Classification Based on Semiadversarial Data Augmentation

Table 3

The performance of different data augmentation modes on the BioBERT model.

ModesAccMicro-RMicro-Micro-F1

Plain73.071.574.773.1
SAR80.378.283.180.8
ERA80.978.282.880.4
LMA84.980.788.184.2
ADA72.171.372.571.9
SemiADA + PI87.788.687.988.2

Four metrics were used to compare the performance differences between the models: accuracy (acc), microprecision (micro-), microrecall (micro-R), and micro-F1 (micro-F1).