Research on the human-AI collaboration framework for news proofreading based on large language models

The digital transformation of journalism has been accelerated by the emergence of Large Language Models (LLMs), bringing profound opportunities and significant challenges to traditional editorial workflows. This paper focuses on manuscript proofreading, a critical function for ensuring news quality and credibility, and proposes a theoretical framework for a Human-AI Collaboration (HAC) system to integrate LLMs into the proofreading process. The framework is built on three pillars: a multilayered system architecture incorporating a Retrieval-Augmented Generation (RAG) knowledge base, a taxonomy of collaboration modes (fully automatic, semi-automatic, and human-led) balancing efficiency and editorial control, and key technical components including prompt engineering and explainable outputs. Theoretical application scenarios are explored, ranging from context-aware error correction to retrieval-augmented fact-checking and compliance screening. A well-designed HAC proofreading system can enhance editorial intelligence, improve consistency, and ultimately strengthen trust in journalism within an increasingly automated media ecosystem.