Bridging the past: a comprehensive survey on the recognition of Japanese historical documents
Japanese historical documents, encompassing formats such as mokkan , kuzushiji manuscripts, and kindai records, represent a continuous documentary heritage linking the Nara period to the modern era. However, the scripts, layouts, and preservation states of these materials pose unique challenges for accessibility, particularly given the scarcity of specialists capable of reading cursive historical scripts such as kuzushiji. This survey provides a comprehensive examination of the recognition and analysis of Japanese historical documents, integrating perspectives from linguistic history, paleography, and computer vision. We review the evolution of the Japanese writing system, outline key challenges in historical document analysis, and present detailed case studies on recognition technologies for mokkan, kuzushiji, and kindai sources. Advances in deep learning, particularly in convolutional and transformer-based models, have improved transcription accuracy and enabled applications in digital humanities, education, archival search, and historical linguistics. We further explore recent trends in pretrained vision–language models, large language model integration, few-shot and zero-shot learning, multimodal reasoning, and self-supervised approaches. Finally, we identify open problems including cross-collection generalization, rare character recognition, interactive tools for humanities scholars, and deeper Natural Language Processing integration, and emphasize the need for sustained collaboration between machine learning and humanities to preserve and revitalize Japan’s written heritage as an accessible resource for global research.
Authors
- Anh Duc Le (ORCID: https://orcid.org/0000-0002-9359-9686)
- Thanh Thi Xuan Lam
- Hieu Duc Thai Tran
- Anh Trac Duc Dinh
- Tho T. Quan (ORCID: https://orcid.org/0000-0003-0467-6254)
- Nhat Tri Ho
Institutions
- Trường ĐH Nguyễn Tất Thành (VN)
- Ho Chi Minh City University of Technology (VN)
Publication Details
- Journal
- International Journal on Digital Libraries
- Published
- 2026-11-19
- DOI
- https://doi.org/10.1007/s00799-026-00456-w
- Primary Topic
- Handwritten Text Recognition Techniques
- Type
- article
- Field-Weighted Citation Impact
- 0.00