Presenter
Mary Ann Tan (@anntan)
Abstract
The German Digital Library (Deutsche Digitale Bibliothek, DDB) aggregates over 65 Million metadata of cultural heritage objects (CHO) from Germany’s archives, libraries, museums, media libraries, and other institutional sectors since its launch in 2014. The collection uses an extension of the Europeana Data Model (DDB-EDM), but no Linked Data access is currently available. GEMEA (German Memory Atlas) fills this gap as the first open knowledge graph over the DDB corpus, built on W3C Semantic Web standards.
Different institutional sectors encode dc: and dcterms: properties with disparate conventions, imposing structural uniformity at the cost of semantic richness. GEMEA embraces this heterogeneity through a data-driven alignment pipeline using MOCHO, a middle ontology that maps DDB-EDM metadata to domain- and modality-specific ontologies via sector-stratified corpus profiling.
GND authority linking connects DDB records to GND Werk entities enabling FRBR-style work grouping of CHO across institutions. Title extraction from catalog strings, a non-trivial task given ISBD punctuation conventions and historical orthography, drives the linking pipeline.
The graph is queryable via a public SPARQL endpoint (QLever), browsable as Linked Data, and accessible to AI agents through a Model Context Protocol (MCP) interface. Source code and data are openly licensed.