READ: Reinforcement-based Adversarial Learning for Text Classification with Limited Labeled Data

TL;DR

This paper introduces READ, a novel reinforcement-based adversarial learning method that leverages unlabeled data to generate synthetic text, significantly enhancing text classification performance with limited labeled data.

Contribution

The paper presents a new semi-supervised approach combining reinforcement learning and adversarial training for text classification with scarce labeled data.

Findings

READ outperforms existing state-of-the-art methods on multiple datasets.

Synthetic text generation improves model generalization.

Reinforcement learning enhances diversity in generated data.

Abstract

Pre-trained transformer models such as BERT have shown massive gains across many text classification tasks. However, these models usually need enormous labeled data to achieve impressive performances. Obtaining labeled data is often expensive and time-consuming, whereas collecting unlabeled data using some heuristics is relatively much cheaper for any task. Therefore, this paper proposes a method that encapsulates reinforcement learning-based text generation and semi-supervised adversarial learning approaches in a novel way to improve the model's performance. Our method READ, Reinforcement-based Adversarial learning, utilizes an unlabeled dataset to generate diverse synthetic text through reinforcement learning, improving the model's generalization capability using adversarial learning. Our experimental results show that READ outperforms the existing state-of-art methods on multiple datasets.