Table 2.

Overview of relation extraction datasets

Dataset name	Published	Data source	Scale/data size	Relation objects	Relation types	Reference
IEPA	2002	PubMed	486 sentences/ ~300 abstracts	Chemicals	Binary interaction	[13]
AIMed	2005	MEDLINE	1955 sentences/ 225 abstracts	Human protein/gene	Binary interaction	[15]
LLL	2005	MEDLINE	77 sentences	Protein/gene in Bacillus subtilis	3 types	[14]
Bioinfer	2007	PubMed	1100 sentences	Protein/gene/RNA and related	68 types	[16]
HPRD50	2007	MEDLINE	145 sentences/ 50 abstracts	Human protein/gene	Binary interaction	[17]
EMU	2010	PubMed	109 abstracts on mutation	Human protein/gene, disease (prostate cancer/breast cancer)	Binary interaction	[55]
MLEE corpus	2012	PubMed	2608 sentences/ 262 abstracts on angiogenesis	Organism, anatomy, molecule types (14 entity types included in 3 major entity types)	Anatomical, molecular, general, planned events (18 events included in 4 major types)	[56]
EU-ADR	2012	MEDLINE	300 abstracts	Drug, disease, and target (gene, protein and sequence variation)	drug–disease, drug–target, target–disease	[57]
ADE corpus	2012	MEDLINE	20 967 sentences/ 2972 documents	Drug, adverse effect and dosage	Drug–adverse effect, drug–dosage	[21]
GAD corpus	2015	PubMed	5329 sentences	Gene and disease	Binary interaction	[58]
PhenoCHF	2015	i2b2 recognizing obesity challenge; PMC	300 discharged summaries; 10 full-texts	Six COPD-related mentions	3 types	[59]
BRONCO	2016	PMC	108 full-texts	Variant, gene, disease, drug and cell-line	Variant with other entities (4 types)	[60]
N-ary	2017	PMC	264 867 sentences	Drug, gene, mutation	Six types (5 pos, 1 neg)	[61]
DDAE dataset	2019	PubMed	521 abstracts (400 for training, 121 for test)	Disease	2 types	[62]
RENET2	2021	MEDLINE; PMC	1000 abstracts; 500 full-texts	Gene and disease	Associated, non-associated and ambiguous	[63]