DR-BiLSTM: Dependent Reading Bidirectional LSTM for Natural Language Inference

TL;DR

This paper introduces DR-BiLSTM, a novel dependent reading bidirectional LSTM model for natural language inference that outperforms previous methods and achieves state-of-the-art results on the Stanford NLI dataset.

Contribution

The paper proposes a dependent reading mechanism within BiLSTM for NLI, along with an ensemble strategy and preprocessing steps, advancing the state-of-the-art performance.

Findings

DR-BiLSTM achieves new state-of-the-art scores on Stanford NLI dataset.

Ensemble strategy significantly improves prediction accuracy.

Preprocessing steps further enhance model performance.

Abstract

We present a novel deep learning architecture to address the natural language inference (NLI) task. Existing approaches mostly rely on simple reading mechanisms for independent encoding of the premise and hypothesis. Instead, we propose a novel dependent reading bidirectional LSTM network (DR-BiLSTM) to efficiently model the relationship between a premise and a hypothesis during encoding and inference. We also introduce a sophisticated ensemble strategy to combine our proposed models, which noticeably improves final predictions. Finally, we demonstrate how the results can be improved further with an additional preprocessing step. Our evaluation shows that DR-BiLSTM obtains the best single model and ensemble model results achieving the new state-of-the-art scores on the Stanford NLI dataset.