Dataset Card for SyNLI

A synthetic NLI datasets from open domain sentences using T5 as data synthesizer. The data can be used to train sentence embedding models.

Data Fields

The data have several fields:

• sent0 : premise as a string
• sent1 : entailment hypothesis as a string
• hard_neg : contradiction hypothesis as a string