CuKa -- Computer-aided cuneiform analysis Cross-repository and cross-domain analysis of cuneiform tablets for collaborative, user-centered operationalization of philological working methods

• Applicants: Professor Dr.-Ing. Gernot A. Fink; Professor Dr. Gerfrid G. W. Müller; Privatdozent Dr. Frank Weichert
• Subject Area: Egyptology and Ancient Near Eastern Studies ; Image and Language Processing, Computer Graphics and Visualisation, Human Computer Interaction, Ubiquitous and Wearable Computing
• Term: from 2018 to 2023
• Project identifier: Deutsche Forschungsgemeinschaft (DFG) - Project number 405966540

The cuneiform script was a writing system that has been used for more than three millennia in the Ancient Near East. As of today, more than 600,000 cuneiform texts have been discovered. In contrast to alphabet characters which are usually represented by a color contrast to the background, cuneiform is a three-dimensional colorless script, which can only be seen under specific lighting due to shadow casting. The challenges lie in the resulting changing shadows, the curvature of the writing surface and the variations of the signs. Therefore, the search in cuneiform texts is currently only possible by means of transcriptions and their indexing, hereby paleographic characteristics are not taken into account. The project is breaking new ground in the retrieval of these documents.The main objective of the research project is to lay the foundations for the provision of a repository/exhibitor spanning both domain-internal (2D-photographs, 3D-scan) and cross-domain search function for finding characters and text passages as well as for operationalizing the analysis. This is to be realized in a web-based, practical and user-centered system for scientific information management for collaborative, systematized philological working methods. The project is intended to make it possible to find text passages quickly, e. g. for collation, by automatically annotating 2D and 3D reproductions/images of the original documents instead of a manual annotation. For the first time, cross-domain analysis tools are provided. Authorized users extend the expertise of the system with each search.For this purpose, modern methods for automated extraction and complex representation of semantic-relevant wedge constellation-specific characteristics are implemented. Furthermore, methods for the automatic identification and probabilistic modeling of line fragment hypotheses are additional objectives. For cuneiform search queries Word spotting-based methods (e. g. for objective character and font comparison) and concepts of Active/Deep Learning are realized. It is also planned to determine a bidirectional mapping between corresponding 2D and 3D datasets with specific cuneiform characteristics.The practical benefit lies in particular in the fast locating of text passages, the search for reference points for character forms (rare or diagnostic characters for dating; comparison characters for damaged characters for text supplements) as well as the cross-document analysis for checking for pen identity or join of fragments and other paleographic questions. The test body comprises over 2,000 own 3D-models of cuneiform scripts in the Hethitologie Portal Mainz and approx. 70,000 photos, more than 100,000 additional photos and 3D-scans are available online.

• DFG Programme: Research data and software (Scientific Library Services and Information Systems)