The project plans to provide guided tours for any given digital archive content, not limited to biographies. The content of archives is often stored in a structured format (e.g., triples), which differs from the textual content we currently use in terms of search and utilization, posing unique challenges. In July 2024, Microsoft released the GraphRag technology and toolkit, which theoretically allows the direct use of triples as reference materials for large language models to perform interactive guided tours. However, the data we obtained from historical figure analysis includes more complex quintuple data (person, event, time, location, and object). Therefore, we plan to modify the GraphRag toolkit to enable it to successfully operate on quintuples, ensuring that different types of knowledge graph formats can be utilized by large language models. This will culminate in a complete interactive guided tour platform that can be integrated into digital cultural platforms and an interactive guided tour app for public use.

This app will resemble a virtual tour guide for archival institutions, using natural language for seamless interaction. It can provide explanations, answer questions, interact, and even ask follow-up questions. For instance, when a digital viewer observes images or 3D renderings of the Jadeite Cabbage, the AI-powered “digital tour guide” can engage in real-time conversations with the viewer. Similarly, when humanities scholars study the life of Li Wei, a Qing dynasty figure, the AI-powered “advisor” can facilitate free-form inquiries and feedback in natural language. It will not only play the role of a data searcher but also analyze the data and provide insights grounded in the archive’s information, rather than relying on the AI’s pre-existing memory.

Such a powerful tool could significantly benefit digital humanities research methods while providing a more accessible medium for the general public to engage with digital humanities.