Publication
This paper introduces an innovative analytical method for visualising research libraries, overcoming the limitations of the assumptions made by their classification systems. The approach combines user loan data with deep mapping techniques to graphically display usage patterns and thematic clusters. Dimensionality reduction is used to visualise the catalogue by book loans, and prompt engineering with large language models is used to describe loan clusters with detailed summaries and titles. This approach was applied to the library collection owned by Bibliotheca Hertziana, a renowned research institute for art history based in Rome. The final output was assessed by a group of experts through interviews supported by an atlas providing statistical information on clusters. This yielded promising results towards a more general framework for visually mapping textual collections and capturing their transformation and usage from an interdisciplinary perspective.