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

2025-01-14

AI Summary

To address the challenges of scarce labeled data and poor generalization in few-shot text classification, this paper proposes a novel framework integrating reinforcement learning–driven text generation with semi-supervised adversarial learning. Specifically, it pioneers the synergistic incorporation of Proximal Policy Optimization (PPO)-guided controllable text generation and Virtual Adversarial Training (VAT) into a consistency-regularized semi-supervised paradigm, leveraging unlabeled data to generate high-fidelity synthetic samples and thereby improving label efficiency and model robustness. Built upon BERT-style pre-trained language models, the method jointly constrains generation quality and discriminative consistency via Monte Carlo sampling and policy gradient optimization. Empirical evaluation across multiple standard benchmarks demonstrates that the approach achieves over 96% of fully supervised BERT performance using only 10% labeled data—substantially outperforming existing state-of-the-art methods.

Technology Category

• Machine Learning: Adversarial Learning & Robustness
• Natural Language Processing: Generation
• Game Theory and Economic Paradigms: Adversarial Learning

Application Category

• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking
• Economics, Online Markets and Human Computation: Humans versus LLMs for data annotation and labeling
• Semantics and Knowledge: Methods to enhance, augment, integrate or synergize semantic models such as knowledge graphs and LLMs

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.

Problem

• Text Classification
• Limited Annotated Data
• BERT

Innovation

• READ method
• Reinforcement Learning
• Semi-supervised Adversarial Learning