A Long-Term Preservation Model of Prompt Digital Repository : Short Paper - iPRES 2025 - Wellington, New Zealand / Te Whanganui-a-Tara, Aotearoa
Abstract
This paper examines the challenges and methodologies associated with the long-term preservation of prompt resources within the domain of artificial intelligence and proposes a storage framework. The article delves into the dynamic dependencies and multimodal characteristics encompassing text, images, and videos associated with prompts. It develops a five-layer storage model comprising: an infrastructure layer (utilizing Docker and hybrid cloud solutions), a data storage layer (incorporating relational and vector databases), a embedding layer (focusing on multimodal embedding), and an application layer (encompassing OAIS processes and API services). The study identifies several technical challenges, including the optimization of storage costs, the safeguarding of privacy, and ensuring version compatibility. Furthermore, it outlines prospective applications in the realm of large medical models, with the objective of providing a systematic solution that ensures the integrity, accessibility, and compliance of high-value prompt resources.Details
- Creators
- Yun-Man Fan; An Fang; Jia-Hui Hu; Chen-Liu Yang; Qian Wang; Lei Wang
- Institutions
- Date
- 2025-09-01 00:00:00 +0000
- Keywords
- digital humanities
- Publication Type
- paper
- License
- http://creativecommons.org/licenses/by-sa/4.0/
- Download
- 336805 bytes