The Getty Provenance Index German Sales as a Large-Scale Language Resource

The presentation reports on research conducted as part of a PhD, exploring the Getty Provenance Index (GPI) as a language resource for AI and computational analysis. As one of the largest multilingual collections of historical art market data, the GPI offers opportunities for linguistic investigation and AI-enabled information retrieval. Using German auction catalogues, BERT-based models were employed to analyse stylistic variation, the language of connoisseurship, specialist terminology, and cataloguers’ use of motif-descriptive vocabulary. Semantic AI pipelines were developed to support multilingual searches across provenance records and improve access to dispersed archival information.

In addition, LLM-based text-mining techniques extracted references to expert attributions and authentications, creating a structured dataset of named experts and their assessments. Together, these case studies illustrate how AI can support the analysis and retrieval of information in large-scale provenance resources.