Assessing The Factual Accuracy of Generated Text

https://arxiv.org/abs/1905.13322

WikiFact

We introduce and release a new large-scale dataset based on Wikipedia and Wikidata to train relation classifiers and end-to-end fact extraction models.