Like most national archives, the National Archives of Latvia (NAL) is a mine of historical documents detailing the intricate past of the Baltic nation. One of the archive’s most important collections is the Minutes of the Riga City Council (1603–1890), which have recently been recognised by UNESCO as part of its Memory of the World Programme. These German-language records cover nearly 300 years of decisions made by the Riga City Council, during periods when Riga was ruled by Poland, Sweden, and later the Russian Empire, making them a valuable source of information about the city and its people.
Until now, this vast collection has only been available in paper form, meaning anyone wanting to view it had to travel to Riga and search through the physical archive. But thanks to the efforts of digitisation expert and archive employee Mairita Lukianska, these important historical records are, for the first time, being made more accessible.
We talked to Mairita to find out more about the project, her Master’s thesis, and about how a Transkribus scholarship has benefited her work.
Just some of the records in the Minutes of the Riga City Council collection. © Latvian State Historical ArchiveA large and important collection
As a professional archivist and a Master’s student in Digital Humanities, Mairita has a strong interest in using technology to preserve historical documents. “I have always been interested in how we can use the latest innovations in information technology to preserve cultural heritage, especially documentary heritage,” she said. Her project to transcribe the Minutes of the Riga City Council is part of her Master’s research on handwritten text recognition (HTR), which aims to make valuable documents such as these easier to digitise, preserve, and access.
And the Minutes of the Riga City Council could not be a better material for this purpose. Consisting of 386 books and more than 263,000 pages, the documents provide a detailed record of the city’s political, economic, and social decisions over nearly three centuries. The collection is such an important part of Latvian history that it was recently included in the UNESCO Memory of the World Programme.
“On 22 March 2022, the Director of the National Archives of Latvia and the Secretary-General of the Latvian National Commission for UNESCO signed a cooperation agreement on the preservation, accessibility, and promotion of the Minutes of the Riga City Council,” Mairita explained. It was this agreement that inspired Mairita to choose the collection for her Master’s thesis and explore how it would be possible to digitally preserve it and make it accessible to everyone.
The volume of the Minutes of the Riga City Council for the years 1643 to 1645. © Latvian State Historical ArchiveTackling the challenges of transcription
But transcribing historical documents, especially handwritten ones, comes with many challenges. The Minutes of the Riga City Council were written in German Kurrent script, a handwriting style that was commonly used during that period. Since the council’s scribes often wrote quickly, their handwriting can sometimes be difficult to read. “One of the biggest challenges is the handwriting. Some handwriting can only be deciphered after careful comparison work,” Mairita explained.
In addition to the handwriting, the documents contain many abbreviations that were typical of administrative texts of the time. These abbreviations are often derived from medieval Latin and don’t have direct modern equivalents, making the transcription even more complex.
Because of these challenges, it would take many years to transcribe the collection by hand. However, Mairita knew that this process could be speeded up dramatically with the help of AI tools. As a first step towards making the Minutes of the Riga City Council fully accessible, she chose a sample of 3,170 pages — 1.2% of the entire collection — and set about finding the best workflow for automatically transcribing it.
The Kurrent script can be tricky to decipher, as shown in this volume from 1710. © Latvian State Historical ArchiveChoosing Transkribus
Ideally, that entire workflow could be carried out with just one tool, saving Mairita valuable time and resources. “Transkribus offers a set of tools for document analysis, including layout and line recognition, tagging, and export in various formats. [It] provides an all-in-one solution in one place, eliminating the need to find tools for each task in the workflow.”
For Mairita, one of the most useful aspects of Transkribus is that it allows users to create their own models based on the handwriting in their specific documents. Given the variety of unique handwriting styles in the Riga City Council minutes, this would be essential for the project. “Training a model was not a difficult challenge on the Transkribus platform. If there were any problems to troubleshoot, I referred to the Transkribus documentation, which is very easy to navigate and well-structured.”
Finally, the fact that Transkribus offers scholarships to early-career researchers was also an important factor in Mairita’s decision. “The scholarship provided financial support for the research. The application form [...] was easy to fill out. I just needed to decide how many pages I wanted to recognise.”
Training a model with Transkribus
Mairita trained her first text recognition model with 109 pages of Ground Truth data, selecting German Kurrent 17th-18th century as a base model. “I [then] applied the model to recognise an additional 200 pages from four different books, [and] submitted the text to the paleographers for editing.”
As Mairita’s professional background is in project management rather than paleography, she worked closely with experts to ensure that these Ground Truth transcriptions were accurate. The handwritten texts were reviewed by Enija Rubina, a senior researcher at the NAL, and Dr. Manfred von Boetticher, an expert in German paleography. “Editing 200 pages took a year,” Mairita says. “It was a time-consuming process but necessary to continue the research.”
Once the corrections were made by the paleographers, Mairita trained a new version of the model, using this expanded collection of Ground Truth. By the end of August 2024, she had used this model to successfully transcribe 3,170 pages — or four books of The Minutes of the Riga City Council.
A transcribed document with the digital text (screenshot taken from a previous version of Transkribus). © Mairita LukianskaDiscovering new details about Riga’s past
During this process, Mairita has come across all sorts of interesting records, covering everything from the Great Northern War and Napoleon’s 1812 campaign, to small details about everyday life. “The books document significant historical events,” Mairita says, “but they also show how closely Riga was connected to European trade and communication networks.”
For example, there is the entry about the Italian who caused a stir in 1644 by bringing his Siamese twin to Riga. Or the Hungarian merchants who were investigated in 1773 for selling an unknown cattle medicine. Then there were the strange regulations that were passed by the Riga City Council during this period, focusing on the illegal import of rhubarb or ruling that hot meals were not allowed to be served at weddings within the city’s boundaries.
While these entries don’t contribute to our understanding of major world events, they do help us to understand how the citizens of Riga lived during this period, what was important to them, and how local and global events shaped their daily lives. Accessing these valuable insights has become much easier thanks to Mairita’s project, particularly for people who lack the paleography skills to read the original documents.
Typical council records from the year 1643. © Latvian State Historical ArchiveLooking to the future
Mairita hopes at some point to make her model public for everyone to use, but wishes to refine it more before doing so. “Currently, the CER for the retrained model on the validation set is 7.50%. I would only make it public if I could improve the model by a few percent more.”
However, as a result of the project, the archive has organised several outreach events, designed to encourage public involvement in the documents. “The NAL have started offering seminars on German handwriting paleography to teach reading skills for 17th-19th century German handwriting. I also led a workshop on how to use [Transkribus] for handwritten documents.”
In addition to these events, Mairita and her colleagues plan to use the digitised documents to test different forms of volunteer involvement in the correction and indexing of the transcribed texts, perhaps paving the way for a future citizen science project.
Advice for future researchers
For students or researchers interested in similar projects, Mairita’s advice is to take advantage of the opportunities available, such as the Transkribus scholarship, and to be prepared for the challenges that come with working on large historical collections. “Working with such a platform provides [an] exceptional experience that can be valuable for your [skillset] and career growth.,” she says.
As she looks ahead, Mairita plans to continue using Transkribus for smaller projects. With the help of new tools and techniques, she hopes to keep uncovering more of Riga’s rich history, ensuring that these important documents remain available for future generations to study, explore, and enjoy.
Thank you for taking the time to talk to us, Mairita!