Pairwise Supervised Contrastive Learning of Sentence Representations

TL;DR

This paper introduces PairSupCon, a contrastive learning method that improves sentence representations by better capturing semantic categories, outperforming previous methods on clustering and similarity tasks.

Contribution

The paper proposes PairSupCon, a novel contrastive learning approach that enhances sentence representations by integrating semantic entailment and contradiction understanding with high-level category encoding.

Findings

Outperforms previous state-of-the-art by 10-13% on clustering tasks.

Achieves 5-6% improvement on semantic textual similarity tasks.

Effectively captures high-level semantic categories in sentence representations.

Abstract

Many recent successes in sentence representation learning have been achieved by simply fine-tuning on the Natural Language Inference (NLI) datasets with triplet loss or siamese loss. Nevertheless, they share a common weakness: sentences in a contradiction pair are not necessarily from different semantic categories. Therefore, optimizing the semantic entailment and contradiction reasoning objective alone is inadequate to capture the high-level semantic structure. The drawback is compounded by the fact that the vanilla siamese or triplet losses only learn from individual sentence pairs or triplets, which often suffer from bad local optima. In this paper, we propose PairSupCon, an instance discrimination based approach aiming to bridge semantic entailment and contradiction understanding with high-level categorical concept encoding. We evaluate PairSupCon on various downstream tasks that involve understanding sentence semantics at different granularities. We outperform the previous state-of-the-art method with $10\%$--$13\%$ averaged improvement on eight clustering tasks, and $5\%$--$6\%$ averaged improvement on seven semantic textual similarity (STS) tasks.