Pharmacovigilance with Transformers: A Framework to Detect Adverse Drug Reactions Using BERT Fine-Tuned with FARM

Table 6

Comparison of results yielded by FARM-BERT with the results yielded by state-of-the-art models on TwiMed corpus.

Models	TwiMed-Twitter			TwiMed-PubMed
	SVM [48]	0.752	0.810	0.778	0.799	0.681	0.728
IAN [48]	0.836	0.813	0.824	0.878	0.738	0.792
CNN-based method [49]	0.739	0.788	0.761	0.849	0.831	0.835
Multichannel CNN [50]	0.738	0.841	0.780	0.861	0.780	0.816
Joint AB-LSTM [51]	0.748	0.856	0.799	0.858	0.852	0.853
MSAM [47]	0.701	0.828	0.754	0.817	0.856	0.831
FARM-BERT	0.831	0.868	0.849	0.952	0.966	0.959