From notes to models: Leveraging LLMs for museum closure data

Our postdoc George A. Wright will present our latest research at the International Conference on Theory and Practice of Digital Libraries (TPDL 2025). In the paper “From Notes to Models: Leveraging LLMs for Museum Closure Data”, we explore how large language models (LLMs) can support researchers in transforming unstructured notes into usable data models. This work is part of the Museum Closure in the UK 2000–2025 project, funded by UKRI-AHRC.

Drawing on a unique corpus of notes documenting the closure of over 500 UK museums, the study introduces a two-stage pipeline, using open-source LLM Llama: (1) model suggestion: the LLM propose schema fragments from chunks of textual notes. (2) model collation: these fragments are combined into a coherent data model that can support further analysis.

The project demonstrates both the potential and the challenges of using LLMs in cultural heritage research, where heterogeneous textual data often remain under-analysed because of time constraints. While the models generated show promise, ensuring consistency and semantic accuracy remains, perhaps inevitably, an open challenge.

➡️ Reference: Wright, G.A., Ballatore, A., Poulovassilis, A., Wood, P.T. (2025) From Notes to Models: Leveraging LLMs for Museum Closure Data, International Conference on Theory and Practice of Digital Libraries (TPDL 2025), Tampere, Finland, 23-26 September 2025.

👉 Below you can find both the poster and the article about this research: