Transcribing Ancient Greek with Transkribus: 3 Case Studies

Of all the world’s ancient cultures, it is perhaps Ancient Greece that has had the most enduring impact on Western civilisation. From philosophy and politics to drama and science, the texts produced during this period laid a foundational framework for centuries of thought and development. For modern scholars, studying these manuscripts provides a unique opportunity to understand the roots of contemporary society, but the work is not without its difficulties. The complexities of the Ancient Greek script can present significant barriers to research.

Fortunately, powerful tools like Transkribus are helping to overcome these barriers.  The platform's capacity for training custom models on complex and historic scripts allows researchers, universities, and archives to digitise and analyse Ancient Greek texts with greater speed and accuracy than ever before. Here are three projects that have successfully used Transkribus to shed new light on historical documents.

Creating a multilingual model for Wüst's lexicon to Aristophanes

The Greek playwright Aristophanes is renowned for his inventive and often unusual linguistic style. His comedies are filled with puns, neologisms, and colloquialisms that can be challenging for even seasoned classicists to interpret. To help scholars, the German classicist Ernst Wüst compiled a comprehensive lexicon in 1910, meticulously cataloguing Aristophanes' unique words and phrases with explanations in German. Recognising how useful this resource was for scholars, Jeff Rusten, a Professor of Classics at Cornell University, sought to create a digital version which could be made freely available online.

The primary challenge was the lexicon's multilingual nature, with entries in Ancient Greek, German, and Latin. A standard text recognition model would struggle to process the three different languages simultaneously. The solution was to use Transkribus to train a specialised, multilingual model capable of accurately transcribing all three languages as they appeared on the page. This bespoke model has not only enabled the creation of a searchable, digital edition of the lexicon but also provides a powerful case study for other projects dealing with multilingual scholarly texts from the 19th and 20th centuries.

Find out more about this project: 

Training a multilingual model in Transkribus — Jeff Rusten

Using Transkribus to transcribe and compare New Testament manuscripts

Transcribing historical manuscripts is a famously time-consuming endeavour, a fact to which Jonathan Wilken at the University of Cambridge can attest. He spent a full year manually transcribing 1700 pages of New Testament manuscripts written in Greek. In traditional textual scholarship, the next step would be to produce a second, independent transcription of each manuscript. This second version, which can be created by the original scholar, a colleague, or a research assistant, is essential for verifying accuracy and identifying potential errors before the critical work of comparing manuscript variants can begin.

Instead of embarking on another lengthy round of manual transcription, Jonathan adopted an innovative approach. He trained a custom Transkribus model capable of reading the Greek scripts used in the manuscripts, and then used this model to automatically produce a second transcription of them. By comparing the model’s output to his own manual work, he could quickly identify and reconcile any discrepancies. This method saved hundreds of hours of transcription work, allowing Jonathan to focus his time on comparison and analysis.

The models Jonathan trained for the project are now publicly available on Transkribus:

• Greek: Ancient Majuscule

• Greek: Medieval and Modern Minuscule

Unlocking the secrets of Karl Benedikt Hase's diary

Karl Benedikt Hase was one of the most significant Greek scholars of the 19th century, but his academic work tells only part of his story. For decades, rumours circulated about a secret personal diary he kept, written almost entirely in Ancient Greek. Though its existence was known, the diary itself remained lost until it was recently rediscovered by William Barton, a Classicist at the University of Innsbruck. This remarkable find promised an candid look into Hase's personal life, the European Philhellenic scene of the 1800s, and his clandestine work forging Ancient Greek manuscripts.

To unlock the diary's contents, William and his team turned to Transkribus. The platform's ability to be trained on Hase’s unique handwriting made it possible to transcribe the diary's handwritten text much more quickly than would have been possible manually, providing a wealth of new information about Hase and his world. The team has also published the transcriptions on a dedicated Transkribus Site, which enables this valuable resource to be shared with the academic community.

“My chief contribution to this project is to read and transcribe the text, then to offer some preliminary studies," William explained. "But publishing everything online means that experts in other fields can take advantage of the work I’ve done and take the research one step further.”

Read the full story of this fascinating discovery: 

Uncovering the diary of Karl Benedikt Hase with START Grant winner William Barton

These three case studies highlight a pivotal shift in classical studies, where automated text recognition is no longer a distant concept but a practical, accessible resource. By addressing specific challenges—such as multilingual documents, difficult handwriting, and the time-consuming nature of transcription verification—this technology provides a practical means for scholars of the Hellenic world to create the accurate, searchable data required for modern analysis.

You can find out more about using Transkribus on our website, on our Help Center, or on our YouTube channel.