Named Entity Recognition for Public Interest Litigation Based on a Deep Contextualized Pretraining Approach

Table 3

Comparison of experimental effects between various models (NER category).

Methods	P	R	F1
CRF [34]	67.7	68.7	68.2
BiLSTM-CRF [34]	69.8	70.6	70.2
WL-BiLSTM-CRF [34]	70.8	71.6	71.2
CNN-BiLSTM-CRF [35]	72.2	72.2	72.2
BERT-FC [36]	75.5	71.7	73.5
BERT-WWM-FC	77.8	72.6	75.1
BERT-WWM-EXT-FC	75.3	75.9	75.6
RoBERTa-FC	74.3	68.3	71.2
RoBERTa-WWM-FC	74.3	69.3	71.7
RoBERTa-WWM-EXT-FC	75.8	70.0	72.8
ALBERT-FC	73.8	64.0	68.6
ALBERT-tiny-FC	68.7	52.9	59.8

The bold values mean the optimal values among all the methods.