Researchers in Germany have unveiled a new artificial intelligence tool capable of identifying individual variations in cuneiform writing, marking a major breakthrough in the study of the ancient Near East.
The system, known as “Palaeographicum”, was developed through a collaboration between the University of Würzburg, the Academy of Sciences and Literature in Mainz, and the Technical University of Dortmund. It is designed to analyse digitised images of cuneiform tablets and compare the unique shapes of wedge-shaped signs across vast collections of texts.
For more than three millennia before the Common Era, civilisations in the Near East recorded administrative, legal, religious, and literary texts on clay tablets using cuneiform writing. Many of these tablets have survived only in fragmentary form, scattered among museums and collections worldwide, making reconstruction an immense challenge for researchers.
The Würzburg-Mainz team has spent decades developing digital tools to support the reconstruction and interpretation of Hittite texts from ancient Anatolia. Their work previously led to the creation of the Hethitologie-Portal Mainz, an online database cataloguing approximately 30,000 known Hittite tablet fragments and related research materials.
More recently, the team introduced digital methods for recording cuneiform signs in three dimensions, followed by the TLHdig search platform in 2023, which enabled searches in both cuneiform and transliterated text.
The latest development, Palaeographicum, expands these capabilities by using AI to identify and compare the distinctive forms of cuneiform signs. The current version provides access to around 70,000 photographs documenting more than five million individual characters.
Researchers say the tool has transformed the field because subtle differences in handwriting can reveal the identity of individual scribes. Although cuneiform signs were impressed into clay using styluses, personal writing styles remain visible in the tablets. Some scribes produced elongated flourishes, while others consistently spaced or angled characters in distinctive ways.
According to Professor Gerfrid Müller, distinguishing these traits manually has traditionally been a slow and difficult process due to the three-dimensional nature of the inscriptions and the varying effects of lighting in photographs.
Professor Daniel Schwemer, head of Ancient Near Eastern Studies at the University of Würzburg, said the AI system has dramatically accelerated research. Tasks that once required several days of visual comparison can now be completed within minutes.
The tool is also expected to improve the dating of Hittite texts. Since most tablets are undated, researchers rely heavily on palaeography—the study of historical handwriting styles—to establish chronology. Changes in writing styles over centuries provide valuable chronological clues.
Development of the AI system remains ongoing. Researchers continue retraining the model using newly annotated material and user feedback from the international Hittitology community.
The long-term goal is to enable the AI to automatically identify individual scribes across large collections of tablets. If successful, scholars believe it could provide unprecedented insight into the professional careers, mobility, and social networks of Hittite scribes.
The foundations of the project were established through the DFG-funded CuKa project (Computer-assisted Cuneiform Analysis), which ran from 2018 to 2023. The AI model underpinning Palaeographicum was developed by Gerfrid Müller together with researchers Christopher Rest and Eugen Rusakov at TU Dortmund University, while philological annotation work was carried out by Turna Somel at the Mainz Academy.
The wider research initiative, “The Corpus of Hittite Festival Rituals”, forms part of Germany’s long-term academies research programme dedicated to preserving and studying global cultural traditions.
Header Image Credit : Daniel Schwemer
Sources : idw