[Retracted] Evidence Prediction Method Based on Sentence Selection for Legal Documents

Table 7

Experimental results based on CJRC 2020 data set (bold is the optimal result).

Model	Ans-F1	Sup-F1	Joint F1
Spider Net (baseline)	70.33	74.61	57.38
BERT-GCN	06.67	76.47	64.23
BERT + CapsNet	77.68	77.42	65.30
BERT + Focal loss	71.2	75.20	58.53
BERT + ResNet2d	76.39	74.52	63.53
BERT + Bi-GRU	69.35	78.25	56.51
Our method	82.56	81.14	70.07