Table 6. Validation results for deep learning models on the 50 K review dataset.

Method	Accuracy	Std. dev.	Class	Precision	Recall	F1-score
RNN	0.953	0.0010	Positive	0.975	0.983	0.979
			Neutral/Mixed	0.744	0.729	0.736
			Negative	0.934	0.888	0.911
GRU	0.954	0.0001	Positive	0.973	0.986	0.979
			Neutral/Mixed	0.764	0.720	0.741
			Negative	0.941	0.890	0.914
LSTM	0.953	0.0001	Positive	0.974	0.984	0.979
			Neutral/Mixed	0.750	0.721	0.736
			Negative	0.938	0.889	0.913
BERT	0.964	0.0057	Positive	0.990	0.980	0.985
			Neutral/Mixed	0.843	0.750	0.794
			Negative	0.944	0.930	0.937