Table 2. Model evaluation.

Model	Category	Precision	Recall	F1	Micro-F1	Macro-F1
BERT-BiLSTM-TextCNN	0	0.9235	0.9085	0.9159	0.9052	0.9143
	1	0.9146	0.9047	0.9096
	2	0.9184	0.9130	0.9157
	3	0.9296	0.9022	0.9157
BERT-BiGRU-TextCNN	0	0.9105	0.9029	0.9067	0.8793	0.8885
	1	0.9062	0.8953	0.9007
	2	0.8883	0.8784	0.8833
	3	0.8542	0.8721	0.8631
BERT-LSTM-TextCNN	0	0.8724	0.8951	0.8836	0.8629	0.8785
	1	0.8843	0.8765	0.8804
	2	0.9023	0.8701	0.8859
	3	0.8528	0.8742	0.8634
BERT-TextCNN	0	0.8749	0.8412	0.8577	0.8598	0.8757
	1	0.8852	0.8685	0.8768
	2	0.8537	0.8821	0.8677
	3	0.9043	0.8957	0.9000
Word2Vec-BiLSTM-TextCNN	0	0.7103	0.7348	0.7223	0.6300	0.6723
	1	0.6892	0.6438	0.6657
	2	0.6719	0.7087	0.6898
	3	0.6155	0.6043	0.6098
Word2Vec-BiGRU-TextCNN	0	0.6361	0.6207	0.6283	0.6075	0.6265
	1	0.6345	0.6531	0.6437
	2	0.6394	0.6112	0.6250
	3	0.6145	0.6026	0.6085
Word2Vec-LSTM-TextCNN	0	0.6581	0.6323	0.6449	0.6189	0.6231
	1	0.6361	0.6129	0.6243
	2	0.6138	0.6199	0.6168
	3	0.6037	0.6084	0.6060
Word2Vec-TextCNN	0	0.6326	0.6109	0.6216	0.6010	0.6145
	1	0.6370	0.6211	0.6289
	2	0.6024	0.6029	0.6026
	3	0.6018	0.6075	0.6046