ContrastNet: A Contrastive Learning Framework for Few-Shot Text Classification

TL;DR

ContrastNet introduces a contrastive learning framework that enhances discriminative text representations and mitigates overfitting in few-shot text classification, achieving superior performance across multiple datasets.

Contribution

The paper proposes ContrastNet, a novel contrastive learning approach that improves few-shot text classification by addressing representation discrimination and overfitting issues.

Findings

Outperforms state-of-the-art models on 8 datasets.

Effectively learns discriminative text representations.

Reduces overfitting through unsupervised contrastive regularization.

Abstract

Few-shot text classification has recently been promoted by the meta-learning paradigm which aims to identify target classes with knowledge transferred from source classes with sets of small tasks named episodes. Despite their success, existing works building their meta-learner based on Prototypical Networks are unsatisfactory in learning discriminative text representations between similar classes, which may lead to contradictions during label prediction. In addition, the tasklevel and instance-level overfitting problems in few-shot text classification caused by a few training examples are not sufficiently tackled. In this work, we propose a contrastive learning framework named ContrastNet to tackle both discriminative representation and overfitting problems in few-shot text classification. ContrastNet learns to pull closer text representations belonging to the same class and push away text representations belonging to different classes, while simultaneously introducing unsupervised contrastive regularization at both task-level and instance-level to prevent overfitting. Experiments on 8 few-shot text classification datasets show that ContrastNet outperforms the current state-of-the-art models.