Table 7. Validation results for deep learning models on the 110 K review dataset.

Method	Accuracy	Std. dev.	Class	Precision	Recall	F1-score
RNN	0.957	0.0006	Positive	0.981	0.982	0.982
			Neutral/Mixed	0.751	0.773	0.762
			Negative	0.930	0.907	0.919
GRU	0.958	0.0002	Positive	0.978	0.985	0.981
			Neutral/Mixed	0.774	0.751	0.763
			Negative	0.930	0.907	0.918
LSTM	0.958	0.0009	Positive	0.979	0.984	0.982
			Neutral/Mixed	0.766	0.754	0.760
			Negative	0.931	0.909	0.920
BERT	0.975	0.0001	Positive	0.989	0.990	0.989
			Neutral/Mixed	0.854	0.846	0.850
			Negative	0.954	0.951	0.952