Transforming Question Answering Datasets Into Natural Language Inference Datasets

摘要

Existing datasets for natural language inference (NLI) have propelled research on language understanding. We propose a new method for automatically deriving NLI datasets from the growing abundance of large-scale question answering datasets.