Hyperbolic sentence representations for solving Textual Entailment

TL;DR

This paper explores the use of hyperbolic space embeddings, specifically the Poincare ball, for textual entailment, demonstrating improved performance over traditional methods on multiple datasets.

Contribution

It introduces a novel approach of embedding sentences in hyperbolic space for textual entailment and provides new datasets for evaluation.

Findings

Outperforms Euclidean baselines on SICK dataset

Second best to Order Embeddings on SNLI dataset

Validates hyperbolic embeddings for hierarchical language data

Abstract

Hyperbolic spaces have proven to be suitable for modeling data of hierarchical nature. As such we use the Poincare ball to embed sentences with the goal of proving how hyperbolic spaces can be used for solving Textual Entailment. To this end, apart from the standard datasets used for evaluating textual entailment, we developed two additional datasets. We evaluate against baselines of various backgrounds, including LSTMs, Order Embeddings and Euclidean Averaging, which comes as a natural counterpart to representing sentences into the Euclidean space. We consistently outperform the baselines on the SICK dataset and are second only to Order Embeddings on the SNLI dataset, for the binary classification version of the entailment task.