Table 7.

Confusion matrix of the deep-learning methods with in-domain training.

Gold-standard method		Prediction method Level-0	Prediction method Level-1	Prediction method Level-2	Prediction method Level-3
BERT_IDPa
	Level-0	2221	131	14	1
	Level-1	137	123	16	0
	Level-2	26	52	50	2
	Level-3	6	6	8	6
RoBERTa_IDPb
	Level-0	2177	176	13	1
	Level-1	128	135	15	0
	Level-2	26	47	52	3
	Level-3	3	6	10	7
XLNET_IDPc
	Level-0	2177	176	13	1
	Level-1	128	130	18	0
	Level-2	26	46	56	2
	Level-3	3	8	10	5

^eBERT_IDP: bidirectional encoder representations from transformers further trained on the in-domain unlabeled corpus.

^bRoBERTa_IDP: robustly optimized bidirectional encoder representations from transformers pretraining approach further trained on the in-domain unlabeled corpus.

^cXLNET_IDP: generalized autoregressive pretraining for language understanding further trained on the in-domain unlabeled corpus.