phenningsson · PyLaia · Published December 11, 2025

Old Swedish Medieval Letters

Text Recognition

Description

HTR model for Old Swedish following a modified TRIDIS (https://hal.science/hal-05008780v1/file/TRIDIS_dataset_HTR.pdf) Transcription Guideline customised for the Swedish National Archives' medieval letter material written in Old Swedish. Persons and places are capitalised (entity capitalisation); punctuation is standardised; consonantal "i" and "u" has been transcribed as "j" and "v", for instance: "iac" --> "jac", “j biscops” --> "i biscops". Abbreviations are expanded and resolved. While this model is not perfect, it serves as a useful starting place to speed up transcription of medieval letters written in Old Swedish. With more training data, accuracy is likely to improve further. The base model that this HTR model is finetuned on is the "German_Gothic_Scripts_14th-16th_century" available on Transkribus.

Try this model

Use this modelOpen in Transkribus
Moderate error rate10.35% CER

Character Error Rate (CER) measures the percentage of characters incorrectly recognised. Lower is better. This model scored 10.35% on its validation set. As a rule of thumb, a CER below 10% is considered good for most handwritten material.

Measured on the model's own validation data. Results on your documents may differ depending on handwriting style, document condition, language, and how closely your material resembles the training data.

Words39,894
Lines2,499
Training Pages144
Model ID450265
Languages
Swedish
Centuries
14th c.15th c.16th c.