Exploiting sequence labeling framework to extract document-level relations from biomedical texts

BMC Bioinformatics

Table 2 Comparison between our method and other state-of-the-art methods on CDR corpus

System	Method	Concept level	P	R	F1
CD-REST [6]	SVM + SVM	Sen + Doc	0.596	0.440	0.507
Gu et al. (2016) [7]	ME + ME	Sen + Doc	0.620	0.551	0.583
Zhou et al. [17]	LSTM-SVM	Sen	0.649	0.493	0.560
Zhou et al. [17]	LSTM-SVM + pp	Sen + Doc	0.556	0.684	0.613
Gu et al. (2017) [8]	CNN + ME	Sen + Doc	0.609	0.595	0.602
Gu et al. (2017) [8]	CNN + ME + pp	Sen + Doc	0.557	0.681	0.613
RPCNN [9]	CNN-RNN	Doc	0.552	0.636	0.591
BRAN [10]	Transformer	Doc	0.499	0.638	0.555
BRAN [10]	Transformer-NER	Doc	0.556	0.708	0.621
Zheng et al. [11]	LSTM-CNN	Doc	0.543	0.659	0.595
Zheng et al. [11]	LSTM-CNN + pp	Doc	0.562	0.680	0.615
Bio-Seq	LSTM-CRF	Doc	0.600	0.675	0.635

ISSN: 1471-2105