Exploiting document graphs for inter sentence relation extraction

Table 8 Examples of errors on the BC5 CDR test set

#	PMID	Chemical-Disease	Golden label	RbSP ^†	SGM ^‡	Type	Effect	Error type
1	2131034	D003561–D020258	CID	NONE	CID	Intra	Better	FN → TP
2	18801087	D000638–D009369	NONE	CID	NONE	Intra	Better	FP → TN
3	44072	C024986–D001145	CID	CID	NONE	Intra	Worse	TP → FN
4	15265979	D005947–D006529	NONE	NONE	CID	Intra	Worse	TN → FP
5	1655018	D000305–D006528	CID	NONE	NONE	Intra	−	FN
6	35781	D010423–D002375	NONE	CID	CID	Intra	−	FP
7	7644931	D017239–D018771	CID	−	CID	Inter	Better	FN → TP
8	10327032	D005472–D008107	NONE	−	CID	Inter	Worse	TN → FP
9	2710809	D001712–D003680	CID	−	−	Inter	−	FN
10	11745287	D016190–D015431	CID	−	NONE	Inter	−	FN
11	10087562	D004280–D008133	NONE	CID	CID	Intra	Worse^∗	FN
12	24464946	D015251–D006331	NONE	−	CID	Inter	Worse^∗	TN → FP

^†The re-implemented intra sentence RbSP model (Can et al. [30]) - without subgraph model in Table 5
^‡subgraph model’s prediction
*Errors due to the imperfect annotation
CID Chemical-induced disease, NONE Unrelated, ‘ −’: Cannot generate path, TP True Positive, TN True Negative, FP False Positive, FN False Negative
Cases where the SBM model gives correct results are highlighted in bold

ISSN: 2041-1480