A Generalised and Adaptable Reinforcement Learning Stopping Method

2025/05/03 by Reem Bin-Hezam, Mark Stevenson, Bin-Hezam, Reem +1 · 1 voice
Computer Science · #FOS: Computer and information sciences #Information Retrieval (cs.IR) #cs.IR

paper · pdf · doi:10.48550/arxiv.2505.01907

arxiv published 2025/05/03 · arxiv updated 2025/07/07

Abstract

This paper presents a Technology Assisted Review (TAR) stopping approach based on Reinforcement Learning (RL). Previous such approaches offered limited control over stopping behaviour, such as fixing the target recall and tradeoff between preferring to maximise recall or cost. These limitations are overcome by introducing a novel RL environment, GRLStop, that allows a single model to be applied to multiple target recalls, balances the recall/cost tradeoff and integrates a classifier. Experiments were carried out on six benchmark datasets (CLEF e-Health datasets 2017-9, TREC Total Recall, TREC Legal and Reuters RCV1) at multiple target recall levels. Results showed that the proposed approach to be effective compared to multiple baselines in addition to offering greater flexibility.

Citations

• Using Chao's Estimator as a Stopping Criterion for Technology-Assisted Review
• Combining Counting Processes and Classification Improves a Stopping Rule for Technology Assisted Review
• Contrastive State Augmentations for Reinforcement Learning-Based Recommender Systems
• TAR on Social Media: A Framework for Online Content Moderation
• Reinforcement learning based recommender systems: A survey
• Providing More Efficient Access To Government Records: A Use Case Involving Application of Machine Learning to Improve FOIA Review for the Deliberative Process Privilege
• Self-Supervised Reinforcement Learning for Recommender Systems
• A Brief Survey of Deep Reinforcement Learning
• Proximal Policy Optimization Algorithms
• Autonomy and Reliability of Continuous Active Learning for Technology-Assisted Review
• Playing Atari with Deep Reinforcement Learning
• A Generalization of Sampling Without Replacement from a Finite Universe
• Reinforcement Learning: An Introduction

Discussions

• Pre-print of #SIGIR2025 paper w/Reem Bin-Hezam “A Generalised and Adaptable Reinforcement Learning Stopping Method” arxiv.org/abs/2505.01907 GRLStop enhances RL-based stopping methods to reduce the ef [bsky, 2 points, 0 comments]

Related

• Reinforcement Learning in Healthcare: A Survey
• The Courage to Stop: Overcoming Sunk Cost Fallacy in Deep Reinforcement Learning
• Reinforcement Learning in Dynamic Treatment Regimes Needs Critical Reexamination
• Reinforcement Learning: An Overview
• Human-Inspired Framework to Accelerate Reinforcement Learning