Stanford Corpus of Implicatives

This is the homepage for the Stanford Corpus of Implicatives presented at NAACL 2019. It is modeled as other NLI datasets, and it includes

• 11K natural language triplets with a premise, hypothesis, and label.
• Human-generated triplets from seed sentences collected from Google Books and the web.
• 90+ implicative constructions.
• Additional 20+ nested implicatives in a extra set of the corpus.
• Joint/disjoint/mismatch
• Nested