. 2023 Oct 24;9:e1639. doi: 10.7717/peerj-cs.1639

Table 2. Comparisons of precision, recall, and $F_{1}$ of the baseline, with EDSE data, bidirectional knowledge distillation method (BKDGEC), fine-tuning, as well as L2R re-ranking.

Bold indicates the highest scores.

Model	QALB-2014				QALB-2015
	Prec.	Recall	$F_{1}$	Prec.	Recall	$F_{1}$
Transformer (baseline)	75.61	55.82	64.22	74.78	60.86	67.10
Transformer + EDSE data	77.14	62.73	69.19	75.36	67.53	71.23
Transformer + EDSE data + BKDGEC	77.91	63.11	69.73	76.17	68.42	72.08
Transformer + EDSE data + BKDGEC + fine-tuning	78.12	63.90	70.29	76.89	69.73	73.13
Transformer + EDSE data + BKDGEC + fine-tuning + L2R re-ranking	78.61	65.59	71.51	78.21	70.28	74.03

Table 2. Comparisons of precision, recall, and F1 of the baseline, with EDSE data, bidirectional knowledge distillation method (BKDGEC), fine-tuning, as well as L2R re-ranking.