An Approach Based on Multilevel Convolution for Sentence-Level Element Extraction of Legal Text

Table 2

The results on RCV1-v2 dataset test dataset. HL represents hamming loss. The symbol “+” denotes that the higher the value is, the better the model performs. The symbol “-” is the opposite.

Models	HL (-)	Micro-P (+)	Micro-R (+)	Micro-F1 (+)
BR	0.0086	0.904	0.816	0.858
CC	0.0087	0.887	0.828	0.857
LP	0.0087	0.896	0.824	0.858
CNN	0.0089	0.922	0.798	0.855
CNN-RNN	0.0085	0.889	0.825	0.856
SGM	0.0082	0.897	0.835	0.864
SU4MLC	0.0077	0.8868	0.8631	0.8748
ML-Reasoner	0.0081	0.912	0.847	0.878
Our model	0.00817	0.8831	0.8682	0.8755