Emil Gunnlaugsson · PyLaia · Published April 30, 2023

Icelandic late 18th century (experimental)

Text Recognition

Description

Model for late 18th century Icelandic – mostly administrative – handwriting. The material used for this model is mostly court documents and letter books from the period 1750 til 1800. It used c.a. 20.000–50.000 words from each hand. It was an experiment in many ways, one was the removal of certain letters like Óó, Öö, Øø and to write them down as only Oo, the same was done for Úú, Üü which became Uu, Ÿÿ, Ýý became Yy etc. One idea behind this was, that if the same was done for Danish 18th century ground truth, more accuracy might be had when combing Icelandic / Danish in one model, which would be desirable since both languages figured heavily in 18th-century Iceland. The material for this particular model is the Icelandic texts that were produced for that combined model. It can in some cases produce fine results if the material in question is applicable. Emil Gunnlaugsson is the creator of the model, but the financing was provided by Centre for Digital Humanities and Arts and the source material by the National Archives of Iceland and Icelandic Manuscript Institute.

Try this model

Use this modelOpen in Transkribus
Very low error rate3.1% CER

Character Error Rate (CER) measures the percentage of characters incorrectly recognised. Lower is better. This model scored 3.1% on its validation set. As a rule of thumb, a CER below 10% is considered good for most handwritten material. This is a larger model trained on diverse material, which generally makes it more robust across different handwriting styles. That said, larger training sets also make it harder to push the CER down further.

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.

Words246,460
Lines32,898
Training Pages532
Model ID51792
Languages
Icelandic