AIDP: A Standards Driven Human-Automation Architecture for Intelligent and Secure Document Indexing

Abstract

Index creation is a cognitively demanding and time-intensive component of scholarly publishing, requiring both mechanical accuracy and expert intellectual judgment. Existing automated indexing tools often fall short of professional standards, while fully manual indexing remains costly and difficult to scale. This paper presents the Advanced Intelligent Document Processor (AIDP), a standards-driven human–AI collaborative architecture designed to support intelligent, secure, and reproducible document indexing. AIDP adopts a human-in-the-loop approach by automating routine mechanical tasks, such as heading detection, alphabetical sorting, deduplication, and locator compression, while reserving semantic decision-making and conceptual organization for human indexers. The system is entirely client-side and web-based, ensuring data privacy for unpublished and sensitive documents, and is explicitly aligned with ISO 999 and the Chinese national standard GB/T 41210-2021. The architecture is implemented through a deterministic, rule-based processing pipeline and evaluated using benchmark documents of varying structural complexity. Experimental results demonstrate high performance, achieving up to 98% accuracy across key indexing metrics while preserving hierarchical integrity. The findings indicate that AIDP offers a practical, scalable solution for professional indexing, bridging traditional indexing expertise with contemporary digital infrastructure.