We trained a BERT language model for Dutch Archaeology, and fine-tuned it to perform Named Entity Recognition for 6 categories of entity: artefacts, materials, time periods, places, contexts and species.

For each document, we extracted all entities, and translated time periods to year ranges. All this information is stored - together with DANS metadata such as author, title, etc - in a JSON file for each document.

This is research output of the PhD research by Alex Brandsen, for the AGNES search engine project.