Enriched Europeana Postcards: Geospatial Metadata Corpus

This dataset contains a pilot corpus of 2,616 digitised postcards retrieved from the Europeana repository and enriched through an AI-based geolocation pipeline developed within the framework of the Time Machine project at EPFL. The pipeline combines multimodal large language models, semantic landmark recognition, and image similarity search to generate geospatial annotations for each postcard, including country, city, and where possible, precise GPS coordinates linked to Wikidata identifiers. Each record is provided as a structured JSON object compliant with the W3C Web Annotation Data Model, and includes the original Europeana metadata alongside the derived geolocation outputs. The dataset is intended to support research in digital cultural heritage, urban history, digital humanities, and AI for geospatial enrichment. The enrichment metadata is released under CC BY 4.0, while original postcard images and metadata retain the licensing conditions specified by Europeana providers.