Published: 26 April 2021| Version 1 | DOI: 10.17632/5p6mzbz2s7.1
Hongyan Zhao


The sentences from NYT corpus of the years 2005 and 2006 are aligned by entities in Freebase for training, while that of the year 2007 are aligned for testing. The dataset has 52 common relations and a special relation NA that there is no relation between entity pair e1 and e2. Overall, there are 522,611 sentences, 281,270 entity pairs and 18,252 relational facts in training data, and 172,448 sentences, 96,678 entity pairs and 1,950 relational facts in test data, respectively.