Berlin tests AI on its colonial paper trail

Berlin’s Ethnologisches Museum put 588,000 pages of its historical archive online through a digitisation project running from 2018 to 2024. The material spans 1830–1947 and includes inventories and collection records. Researchers can consult the documents without travelling to the museum.[2]

Its 2025–27 AI pilot addresses another obstacle: old German handwriting and a German-only record. The museum is testing transcription, translation into several languages and data visualisation, focusing on documents about naval activities in colonial German New Guinea.[1]

The archive records a collecting system as well as individual acquisitions. In 1899, a parliamentary resolution made the museum one of Berlin’s central collection points for natural and cultural objects from German colonies. Documentation of acquisitions, exchanges and payments allows researchers to reconstruct the networks through which objects moved.[2]

The three AI tasks serve different purposes. Transcription makes handwriting readable as text; translation widens who can read it; visualisation can organise recorded connections for investigation. The pilot concerns records associated with colonial military campaigns and objects acquired in that context.[1]

The earlier digitisation covered 1,603 historical documents. That document count and the much larger page count describe the same archive at different scales. Locating a relevant file is only the beginning of following its correspondence and entries.[2]

One central question in the pilot is how much human work is needed to turn automated processing into correct, usable research material. That is the practical bridge between putting a collection online and opening its history to scrutiny.[1]

Report an error or suggest a correction