Table 4.

Results based on BiGRU-CRF in three tests with different pretraining models (%).

Model	Entity	Test 1			Test 2			Test 3
		Precision (%)	Recall (%)	F1 (%)	Precision (%)	Recall (%)	F1 (%)	Precision (%)	Recall (%)	F1 (%)
BERT	Cheating	93.42	93.42	93.42	98.63	94.74	96.64	98.63	94.74	96.64
	Mixed	100.00	99.85	99.92	100.00	100.00	100	100.00	100.00	100
	Out-listed	96.66	98.43	97.54	96.29	98.30	97.28	96.29	98.30	97.28
	Not fresh	92.06	92.06	92.06	89.23	92.06	90.62	89.23	92.06	90.62
	Frozen	82.11	76.52	79.22	83.61	77.27	80.31	83.61	77.27	80.31
	Spoilage	91.87	96.58	94.17	94.21	97.44	95.8	94.21	97.44	95.8
	Over-loading	99.32	98.65	98.99	99.55	99.21	99.38	99.55	99.21	99.38
	Deep-processing	83.19	85.34	84.26	84.42	85.63	85.02	84.42	85.63	85.02
ALBERT	Cheating	68.57	63.16	65.75	77.61	68.42	72.73	77.61	68.42	72.73
	Mixed	96.54	97.57	97.05	96.56	98.18	97.36	96.56	98.18	97.36
	Out-listed	86.15	91.90	88.93	87.69	92.16	89.87	87.69	92.16	89.87
	Not fresh	73.85	76.19	75	69.12	74.60	71.76	69.12	74.60	71.76
	Frozen	76.61	71.97	74.22	77.24	71.97	74.51	77.24	71.97	74.51
	Spoilage	75.71	90.60	82.49	72.03	88.03	79.23	72.03	88.03	79.23
	Over-loading	92.28	92.59	92.44	91.51	91.92	91.71	91.51	91.92	91.71
	Deep-processing	63.99	70.98	67.3	64.25	68.68	66.39	64.25	68.68	66.39
RoBERTa	Cheating	95.95	93.42	94.67	93.67	97.37	95.48	93.67	97.37	95.48
	Mixed	100.00	99.85	99.92	100.00	100.00	100	100.00	100.00	100
	Out-listed	97.30	99.08	98.19	97.16	98.56	97.86	97.16	98.56	97.86
	Not fresh	95.08	92.06	93.55	93.55	92.06	92.80	93.55	92.06	92.80
	Frozen	83.61	77.27	80.31	85.48	80.30	82.81	85.48	80.30	82.81
	Spoilage	92.62	96.58	94.56	91.06	95.73	93.33	91.06	95.73	93.33
	Over-loading	99.22	99.33	99.27	99.44	99.10	99.27	99.44	99.10	99.27
	Deep-processing	80.51	81.90	81.2	84.01	83.05	83.53	84.01	83.05	83.53