TINA: Textual Inference with Negation Augmentation

Transformer-based language models achieve state-of-the-art results on several natural language processing tasks.One of these is textual entailment, i.e., the task of determining whether a premise logically entails a hypothesis.However, the models perform poorly on this task when the examples contain negations.In this paper, we propose a new definition of textual entailment that captures also negation.This allows us to develop TINA (Textual Inference with Negation Augmentation), a principled technique for negated data augmentation that can be combined with the unlikelihood loss function.Our experiments with different transformer-based models show that our method can significantly improve the performance of the models on textual entailment datasets with negation -without sacrificing performance on datasets without negation.Proof.Assume that there exist A and B such that A ▷ B and ¬A ▷ B. Then P (B|A) > P (B) and P (B|¬A) > P (B).Hence, we have: P (B) = P (A) × P (B|A) + P (¬A) × P (B|¬A) ⇒ P (B) > P (A) × P (B) + P (¬A) × P (B) ⇒ P (B) > P (B) This is a contradiction, which proves the claim.Proposition A.4.For all events A and B, if A ▷ B then A ̸ ▷ ¬B.