Skip to main content
. 2022 Jan 14;9:812735. doi: 10.3389/fpubh.2021.812735

Table 9.

Comparison of proposed transformer based model accuracy with baseline models.

Model dataset COVIDSenti COVIDSenti_A COVIDSenti_B COVIDSenti_C
Existing models/Accuracy
distilBERT 93.09% 93.07% 92.09% 92.06%
BERT 94.08% 94.01% 93.07% 92.02%
XLNET 93.03% 92.04% 91.04% 92.00%
ALBERT 92.09% 91.04% 92.00% 91.01%
Proposed model/Accuracy
Multi-depth DistilBERT 96.66% 95.22% 94.33% 93.88%
Gain 2.58% 1.21% 1.26% 1.82%

The bold ones represent better results.