A new research initiative is bringing artificial intelligence (AI) and archaeology together to unlock the mysteries of the ancient Hittite civilization. Researchers from Julius-Maximilians-Universität Würzburg (JMU) are leading the project, which is reshaping how experts analyze and connect thousands of clay tablet fragments left behind by one of Anatolia's most powerful Bronze Age empires.
The project, titled "Digital Pathways to the Hittite World," aims to modernize the Hethitologie-Portal Mainz (HPM), a long-standing digital resource for Hittite studies. By integrating tools from AI, archaeology and philology, the team plans to expand the portal into a cutting-edge platform for research on the Hittite world.
Professor Daniel Schwemer, head of Ancient Near Eastern Studies at JMU, is directing the project along with collaborators Stephanie Döpper, Martin Gruber and Andreas Schachner of the German Archaeological Institute in Istanbul.
Digital infrastructure and AI support are provided by JMU's Center for Philology and Digitality (ZPD) and the Center for Artificial Intelligence and Data Science (CAIDAS), under the guidance of Professor Andreas Hotho.
One of the project's priorities is improving data quality. Since excavations at Boğazköy-Ḫattuša began in 1906, records have varied in consistency. Döpper said differences in how excavation sites were documented over the decades make structured searches difficult.
The team is addressing this by cleaning metadata, standardizing location names and improving how fragments are described and linked.
In addition to refining existing data, the project introduces AI-driven language tools. A German-trained large language model known as LLäMmlein is being adapted to learn Hittite.
This model will help researchers reconstruct fragmentary texts by predicting likely continuations and identifying linguistic patterns, speeding up the assembly of incomplete tablets.
Another major focus is resolving inconsistencies in older field records. Researchers are reviewing original excavation logs to standardize terminology and ensure accurate mapping of where tablets and objects were found.
The team is also improving classification methods for ceramic sherds, moving away from unreliable features like preserved length toward more accurate digital typology systems.
The project builds on nearly 25 years of work on the HPM, which began in 2001 with support from the Deutsche Forschungsgemeinschaft.
A new phase of development, backed by VolkswagenStiftung, will include Linked Open Data standards, geolocation of finds and enhanced shape recognition for Hittite artifacts.
The Hittites, who ruled much of Anatolia between 1650 and 1200 BCE, left behind thousands of cuneiform tablets. By linking texts, objects and archaeological context through digital tools, the project aims to open new paths of discovery across disciplines.