e-SNLI: Natural Language Inference with Natural Language Explanations

TL;DR

This paper introduces e-SNLI, an extended dataset with human-annotated natural language explanations for NLI, enabling models to produce interpretable decisions and improve robustness and transferability.

Contribution

It provides a new dataset with explanations and demonstrates models that incorporate explanations for better interpretability and performance in natural language inference tasks.

Findings

Models trained with explanations produce more interpretable outputs.

The dataset improves transfer learning to out-of-domain NLI datasets.

Explanations enhance sentence representation quality.

Abstract

In order for machine learning to garner widespread public adoption, models must be able to provide interpretable and robust explanations for their decisions, as well as learn from human-provided explanations at train time. In this work, we extend the Stanford Natural Language Inference dataset with an additional layer of human-annotated natural language explanations of the entailment relations. We further implement models that incorporate these explanations into their training process and output them at test time. We show how our corpus of explanations, which we call e-SNLI, can be used for various goals, such as obtaining full sentence justifications of a model's decisions, improving universal sentence representations and transferring to out-of-domain NLI datasets. Our dataset thus opens up a range of research directions for using natural language explanations, both for improving models and for asserting their trust.