CLEAR: Contrastive Learning for Sentence Representation

TL;DR

CLEAR introduces a contrastive learning approach for sentence representations using diverse augmentation strategies, leading to improved performance on benchmarks by learning noise-invariant sentence embeddings.

Contribution

The paper proposes a novel contrastive learning method with multiple sentence augmentation techniques for better sentence-level representations.

Findings

Outperforms existing methods on SentEval and GLUE benchmarks

Different augmentations improve performance on various tasks

Enhances noise robustness of sentence embeddings

Abstract

Pre-trained language models have proven their unique powers in capturing implicit language features. However, most pre-training approaches focus on the word-level training objective, while sentence-level objectives are rarely studied. In this paper, we propose Contrastive LEArning for sentence Representation (CLEAR), which employs multiple sentence-level augmentation strategies in order to learn a noise-invariant sentence representation. These augmentations include word and span deletion, reordering, and substitution. Furthermore, we investigate the key reasons that make contrastive learning effective through numerous experiments. We observe that different sentence augmentations during pre-training lead to different performance improvements on various downstream tasks. Our approach is shown to outperform multiple existing methods on both SentEval and GLUE benchmarks.