数字图书馆论坛2026，Vol.22Issue(3)：31-39,9.DOI:10.3772/j.issn.1673-2286.2026.03.004
基于生成式语义解耦的古籍视觉知识元重构与可信归档研究
Research on the Reconstruction and Trustworthy Archiving of Ancient Book Visual Knowledge Elements Based on Generative Semantic Decoupling
摘要
Abstract
The digitization of ancient books is moving toward higher quality and intelligence.However,the current whole-page scan as the basic management unit suffers from granularity mismatch,weak semantic association,and ambiguous evidence provenance.Generative Artificial Intelligence is shifting ancient book image processing from text recognition toward semantic understanding and visual reconstruction.In response,this study introduces the concept of visual knowledge unit with its formal definition and classification.Based on visual knowledge unit theory,a three-layer intelligent archiving architecture is constructed and integrated with the Open Archival Information System and relevant Chinese industry standards,aiming to achieve immutable evidence,controllable derivation,and verifiable processes.This study further presents a reconstruction workflow for semantic decoupling,followed by application scenarios and preliminary validation using ancient books from the Nanjing Library collection.The work provides a practical framework and pathway for moving ancient book resources from image-based digitization to semantic digitization.关键词
视觉知识元/古籍数字化/生成式人工智能/语义解耦/OAIS/知识组织/数字人文Key words
Visual Knowledge Unit/Ancient Book Digitization/GenAI/Semantic Decoupling/OAIS/Knowledge Organization/Digital Humanities分类
社会科学引用本文复制引用
刘丽娜,蔡茜..基于生成式语义解耦的古籍视觉知识元重构与可信归档研究[J].数字图书馆论坛,2026,22(3):31-39,9.基金项目
本研究得到2025年度江苏省图书馆学会一般课题"智慧图书馆的技术赋能与创新发展研究"(编号:25YB074)资助. （编号:25YB074）
