Named Entity Recognition in Medieval Documents

This page lists the resources for named entity recognition in medieval documents that were developed in ​the AHISTO project.

• A human-annotated dataset for language modeling and named entity recognition in medieval documents
• Large language models for named entity recognition in medieval documents:
  • ​Model L: our largest and most accurate model
  • ​Model S: a smaller and more efficient variant of ​Model L, used in ​our named entity recognition tool
  • ​Model TDS1 and ​Model TDS2: two smaller models trained on data of different size as a part of our ablation study
  • ​Model CI: a smaller model trained with a loss function that does not address class imbalance as a part of our ablation study
• ​Named entity experiments: experimental code for named entity search and recognition in medieval documents
• ​Named entity recognition tool: a command-line tool for named entity recognition in medieval documents, see also an online demo (in Czech)

Citing

An article describing our dataset is currently under review. Preprint is available ​on request.