Using AI to Bring Art Research to Life
This resource introduces the use of Large Language Models for transforming unstructured PDF art catalogues into structured digital resources and discusses how AI-driven workflows enhance art historical research and public engagement using the Corpus Rubenianum (a series of over 40 volumes on Rubens) as a case study, highlighting techniques such as LLM document parsing, Pydantic schema-constrained extraction, interactive timelines and network graph visualizations, and open-set computer vision detectors. Particular emphasis is placed on extracting artwork provenance data, resolving abbreviated bibliographic references, and identifying iconographic characters for interactive digital overlays, while also underlining the broader potential of modern AI tools to unlock legacy scholarly publications for interactive exploration and cultural heritage applications.
This presentation is part of the Digital Humanities After Hours lecture series, hosted by the Centre for Digital Humanities and Arts (DARIAH-IS/CDHA, https://www.mshl.is) in Autumn 2025.
Learning Outcomes
After studying the resource, learners should be able to:
- convert unstructured PDF catalogues into structured JSON datasets by combining Large Language Models with predefined metadata schemas;
- construct interactive network graphs and historical timelines using extracted entities, including provenance chains, artwork classifications and relational metadata;
- deploy Visual Language Models and open-set object detection tools to identify and locate specific mythological figures and visual details in paintings from text prompts;
- evaluate re-use strategies for legacy print scholarship within cultural heritage contexts, from digital database enrichment to augmented reality museum experiences.
