On Isotropy, Contextualization and Learning Dynamics of Contrastive-based Sentence Representation Learning

TL;DR

This paper investigates why contrastive learning improves sentence representations by analyzing its effects on isotropy, contextualization, and learning dynamics, revealing geometric and semantic properties of the embedding space.

Contribution

It provides a geometric interpretation of contrastive SRL, highlighting how it promotes isotropy, intra-sentence similarity, and the effects of training parameters on learning dynamics.

Findings

Contrastive learning enhances isotropy in sentence embeddings.

Tokens within the same sentence converge in semantic space.

Training dynamics are influenced by temperature, batch size, and pooling methods.

Abstract

Incorporating contrastive learning objectives in sentence representation learning (SRL) has yielded significant improvements on many sentence-level NLP tasks. However, it is not well understood why contrastive learning works for learning sentence-level semantics. In this paper, we aim to help guide future designs of sentence representation learning methods by taking a closer look at contrastive SRL through the lens of isotropy, contextualization and learning dynamics. We interpret its successes through the geometry of the representation shifts and show that contrastive learning brings isotropy, and drives high intra-sentence similarity: when in the same sentence, tokens converge to similar positions in the semantic space. We also find that what we formalize as "spurious contextualization" is mitigated for semantically meaningful tokens, while augmented for functional ones. We find that the embedding space is directed towards the origin during training, with more areas now better defined. We ablate these findings by observing the learning dynamics with different training temperatures, batch sizes and pooling methods.