Multipath Cross Graph Convolution for Knowledge Representation Learning

Table 1

Comparison of characteristics of different datasets.

Dataset	#Ent	#Rel	#Train	#Valid	#Test
FB15k [24]	14,951	1,345	483,142	50,000	59,071
WN18RR [29]	40,943	18	141,442	2,500	2,500
YAGO43k [30]	42,335	37	331,687	29,599	29,593
FB15kET	14,951	3,851	136,618	15,749	15,780
WN18R	40,501	10	110,341	2,100	2,100
YAGO43kET	41,723	45,182	375,853	42,739	42,750
FB15kTRT	3,851	1,345	2,015,338	—	—
YAGO43kTRT	45,128	37	1,727,707	—	—
WN18-R	6,501	5	90,341	—	—