Pathways and Practices for AI-Powered Digital Archives and Intelligent Services

Abstract

Given the practical challenges of digital archives, such as resource heterogeneity, insufficient semantic utilization, and limited service capabilities, this research constructs an integrated AI-enabled digital archive system encompassing a "data layer—model layer—service layer—operation and maintenance layer." This system unifies the modeling and processing of archival metadata, full-text content, and user behavior, building a knowledge graph and semantic vector framework. It develops semantic retrieval and multi-dimensional navigation algorithms combining deep text encoding, KG embedding, and learning-based ranking, and then utilizes retrieval enhancement to generate intelligent question-answering and personalized recommendation models. Starting from the system level, it constructs a closed loop of intelligent service and manual review through service orchestration and human-machine collaboration. Offline evaluations and simulated online trials conducted on pilot platforms demonstrate that the proposed solution outperforms traditional methods in terms of retrieval accuracy, question-answering quality, recommendation click-through rate, and processing efficiency, while ensuring that tail latency and the proportion of manual intervention remain within a controllable range. This system provides a scalable technical approach and practical reference example for the intelligent upgrading of digital archives and information services.