A large annotated corpus for learning natural language inference

TL;DR

This paper introduces a large-scale, human-annotated corpus for natural language inference, enabling significant improvements in semantic understanding models and setting new benchmarks in the field.

Contribution

It presents the Stanford NLI corpus, a 570,000 sentence pair dataset, significantly larger than previous resources, facilitating advancements in entailment and contradiction modeling.

Findings

Lexicalized classifiers outperform some sophisticated models.

Neural network models perform competitively on NLI tasks.

The dataset enables scalable and improved semantic inference.

Abstract

Understanding entailment and contradiction is fundamental to understanding natural language, and inference about entailment and contradiction is a valuable testing ground for the development of semantic representations. However, machine learning research in this area has been dramatically limited by the lack of large-scale resources. To address this, we introduce the Stanford Natural Language Inference corpus, a new, freely available collection of labeled sentence pairs, written by humans doing a novel grounded task based on image captioning. At 570K pairs, it is two orders of magnitude larger than all other resources of its type. This increase in scale allows lexicalized classifiers to outperform some sophisticated existing entailment models, and it allows a neural network-based model to perform competitively on natural language inference benchmarks for the first time.