Non Departmental Units
Projects & Collaborations

GraphAfghanica: Building a Federated Knowledge Infrastructure for Afghan Historical Images
Research Project | 2 Project Members
GraphAfghanica aims to develop a federated digital humanities infrastructure for the study of historical photographs of Afghanistan from the 1970s, a decade marked by profound social and political transformation. Afghan visual heritage from this period is widely dispersed across international institutions, private collections, and local archives, where it remains unevenly described, difficult to access, and rarely studied in a connected manner. Focusing initially on two major collections, the Bibliotheca Afghanica (BA) at the University Library of Basel and the EWA-76 ethnographic collection preserved in the Józef Burszta Digital archive in Poland, the project addresses this fragmentation by interlinking distributed photographic holdings through a provenance-aware
knowledge graph (KG). Rather than centralizing data, GraphAfghanica enables cross-collection exploration while preserving institutional custodianship and data sovereignty. A dedicated web-based application will provide scholars and the public with integrated access to images, metadata, and relationships across repositories, supporting chronological, spatial, and thematic inquiry into Afghanistan’s visual history.
Methodologically, the project integrates knowledge representation with interpretable computer vision and controlled multilingual description generation within a transparent, provenance-rich framework. Photographs will be enriched through image classification, similarity detection, and GraphRAG-based description generation, with all computational outputs explicitly documented, confidence-scored, and open to scholarly scrutiny. By embedding these methods directly into the KG, GraphAfghanica enables historically meaningful analyses of visual patterns, such as transformations of urban environments and changes in women’s visibility in public space. All software components, ontologies, and workflows will be released as open-source resources, ensuring transferability beyond the Afghan case study. GraphAfghanica thus delivers both a sustainable research infrastructure and a reusable methodological framework for the responsible application of AI-assisted approaches to historical visual archives, advancing digital humanities scholarship while supporting ethical engagement with culturally sensitive heritage.





