The National Library of Norway has organized a seminar on the Hugin-Munin research project and the development of the NorHand corpus and model for document recognition in Norwegian.

The project Hugin-Munin has focused on developing AI models for recognising historical Norwegian handwriting and the work has resulted in several available state-of-the-art models and an open dataset with over 10,000 manually annotated letter and diary pages from the 18th and 19th centuries. The dataset can be used for training and evaluation of handwriting recognition on historical Norwegian letters. It is available on HuggingFace.

The seminar has provided insight into both results from the Hugin-Munin project itself, as well as how these results have contributed to other research projects and how you can benefit from the Hugin-Munin resources in your own research. The target group was researchers in automatic handwriting recognition and anyone who works with handwritten documents in Norwegian and who uses or wants to use automatic handwriting recognition.

Christopher Kermorvant presented TEKLIA's contribution to the project and more broadly to the open source ecosystem for historical document recognition.