How the Hanse.Quelle.Lesen! project made Hanseatic records accessible through Citizen Science
The Hanse.Quelle.Lesen! project is a collaboration between the Research Centre for Hanse and Baltic History (FGHO), based at the European Hansemuseum Lübeck, and the Archive of the Hanseatic City of Lübeck. Launched in 2020, the initiative aims to digitise the Hanserezesse (official meeting minutes of Hanse towns) and other Hanseatic records from the 14th to 17th centuries. By converting these handwritten records into searchable text with the help of Citizen Science, the project seeks to make the political and economic world of the German Hanse accessible to both professional researchers and the general public.
Manuela Nitsch, Citizen Science Manager at the FGHO and Dr Angela Huang, Head of the FGHO. © Olaf Malzahn
The Challenge: The "Ground Truth" bottleneck
While modern AI has the potential to "read" historical documents and transcribe them at scale, it cannot do so accurately without a foundation of high-quality training data, known as Ground Truth. The project faced a massive volume of archival material: thousands of pages of complex legal and trade records written in Middle Low German and early modern scripts such as Kurrent. Depending on the complexity of the handwriting, each page takes around an hour for a trained paleographer to decipher manually.
When the project started back in 2017, far fewer public models were available, particularly for Middle Low German.“Initially, we wanted to access [the documents] ourselves using AI, but we didn't get very far because you always have to create your own training data first [...]. Then the idea arose to do it together with the public,” - Angela Huang, head of the FGHO
To use volunteers for manual transcription, the team needed a platform that was easy to navigate, regardless of a user's technical experience. Since many volunteers work from home, the project also required a centralised, web-based hub accessible from anywhere.
The FGHO decided to use volunteers to manually transcribe documents and create Ground Truth data to train AI text recognition models. © Nadine Sommer
The Solution: Transkribus as a central hub for Citizen Science
The project team selected Transkribus as the ideal solution because it integrates the entire workflow into a single, user-friendly interface. The team could use the platform not only to create Ground Truth but also to train AI models directly, without the need to move data between different systems.
Transkribus’ intuitive design enabled citizen scientists to perform the essential task of manually transcribing documents. The web-based platform made it easy for volunteers to access and transcribe documents, even from the comfort of their own homes.
“When new citizen scientists join the project we’ll meet with them one-to-one online to give them basic information about the project and how to use Transkribus. Since many of our volunteers are older or not as well-practiced with digital tools, a changed user interface, for example, can cause confusion. So I also explain new Transkribus features or changes during our weekly meetings,” - Manuela Nitsch, Citizen Science Manager, FGHO
Once the pages have been transcribed by the volunteers, the data is used within Transkribus to train custom text recognition models capable of reading the specific handwriting used in the Hanseatic documents. As more Ground Truth is created, these models can then be retrained to improve their accuracy.
The transcribed documents are uploaded onto the team's Transkribus Site, making them digitally accessible to the general public. © Transkribus
Benefits: Speed, precision, and social engagement
Using Transkribus had four key benefits for the project:
- Streamlined data creation: By giving volunteers the tools to easily transcribe documents online, the FGHO team were able to produce Ground Truth data on a scale that had previously been impossible. “Without Ground Truth for our documents, we wouldn't have been able to train tailored models for the "Hanserezesse" documents, as producing this amount of Ground Truth would not have been possible for our small team,” - Manuela Nitsch
- Technical accuracy: Thanks to the extensive Ground Truth provided by volunteers, the project developed custom AI models that achieved a Character Error Rate (CER) of just 5%, outperforming even the high-quality Super Models on documents of this type. “These models can now be applied to the many Hanseatic sources still untranscribed. While transferability to similar sources isn't guaranteed, the models provide a strong foundation for fine-tuning on related material,” - Manuela Nitsch
- Accessible resources: Once a document has been fully transcribed and post-processed by the FGHO team, it is made publicly accessible on Transkribus Sites. “We prioritise transcribing the "Rezesse" of the later Hanseatic period to make them available for scientists and the interested public alike,” - Manuela Nitsch
- Human impact: Beyond the data, the project offers a significant social benefit. Volunteers take an active role in preserving history, gaining a sense of ownership over our collective past while improving their paleographical skills in a supportive environment. Some volunteers have been with the project since 2020, and have transcribed hundreds of pages in that time. “It's a really great feeling for me to know that the transcription work isn't just gathering dust on a shelf, but is actually actively in demand,” - Kristina Russ, citizen scientist for Hanse.Quelle.Lesen!
Volunteers such as Kristina Russ attend weekly online meetings, where they are updated on the project and the Transkribus platform. © Nadine Sommer
Conclusion: A new model for historical research
The Hanse.Quelle.Lesen! project demonstrates that the "bottleneck" of manual transcription is not a barrier, but an opportunity for engagement. By combining the technical power of Transkribus with the dedication of the public, the FGHO has turned inaccessible historical records into a truly accessible digital archive. This collaboration proves that when people have the right tools and institutional support, they don't just help process data; they become an integral part of preserving it for the future.
Interested in using Transkribus for your Citizen Science project?
Transkribus has been used by archives and research institutions around the world to engage volunteers in historical research. If you want to find out more about how Transkribus could benefit your Citizen Science project, book a consultation with our advisors and get tailored advice on workflow design, model selection, and custom plans.
