We present OldBERTur, a Named Entity Recognition (NER) model for Old Icelandic available in two variations, one for normalised texts, and one for diplomatic texts. Using a BERT-based model architecture, we fine-tune an existing BERT language model, and due to training data scarcity, we employ multiple training configurations, including pre-training domain adaptation, sentence-level data resampling, and modern Icelandic data augmentation; achieving a 93 F1 score for normalised texts, and 79 for diplomatic texts. We find that additional training configurations, such as resampling entity-annotated Old Icelandic texts, significantly improve performance in low-resource settings, while the effectiveness of added training configurations diminishes as the available training data increases. Our models can be used to automatically identify and classify person and location names in texts sourced from the rich Icelandic medieval literary tradition. Our models, along with their data and code, are made publicly available to allow for reuse and future research into medieval Scandinavian NLP and beyond.

Our approach primarily centers on feature-based systems, where a diverse array of features pertinent to the text’s linguistic attributes is extracted. Alongside those, we incorporate token-level probabilistic features which are fed into a Bidirectional Long Short-Term Memory (BiLSTM) model. Both resulting feature arrays are concatenated and fed into our final prediction model. Our method under-performed compared to the baseline, despite the fact that previous attempts by others have successfully used linguistic features for the purpose of discerning machine-generated text. We conclude that our examined subset of linguistically motivated features alongside probabilistic features was not able to contribute almost any performance at all to a hybrid classifier of human and machine texts.