Documentary heritage is a valuable testament to human history and culture, but its vast volume, complexity, and specialized nature hide much knowledge. Current technological limitations restrict the development and utilization of this heritage. This study proposes a dual-track activation model with LLMs as the core, using a low-code approach to learn from and impart knowledge from vast documentary heritage through personalized services, aiming to prevent it from being sealed away in the 'dusty archives'.
The pattern track, based on digital humanities and knowledge management, constructs an "Extraction-Reconstruction-Service" model to extract key knowledge elements, reconstruct them into a structured system, and integrate user needs into an interactive service platform.

Beneath the pattern track, the capability track establishes a "self-acquisition, self-memory, self-application" technical path based on LLMs, enabling continuous optimization while executing the steps of the pattern track. First, LLMs utilize controlled and open knowledge extraction methods to acquire knowledge elements from raw resources. Next, knowledge embedding techniques are used to embed the knowledge from documentary heritage into the LLM in a vectorized form, allowing the LLM to "remember" domain-specific knowledge. Finally, the LLM combines existing knowledge with new memories to provide personalized and refined services, such as search, Q&A, and recommendations.
Based on the dual-track activation model, this study developed a Suzhou Silk Pattern Service Platform, using the "Silk Archives of Suzhou, China" from the Memory of the World registry as the primary resource.

Authors:
Chi Jin, Renmin University of China, China
Anrunze Li, Renmin University of China, China
Li Niu, Renmin University of China, China
Rundong Hu, Renmin University of China, China
Wancheng Yang, Renmin University of China, China

About the Presenter(s)
Ms. Jin Chi is currently a Ph.D. candidate at the School of Information Resource Management, Renmin University of China. Ms. Jin's research focuses on digital humanities, the digitization of cultural heritage, and knowledge services.

See this presentation on the full schedule – Monday Schedule

Bookmark