SBIR-STTR Award

This Small Business Innovation Research (SBIR) Phase I project will create new software for automatically creating textured, 3D digital models of cuneiform inscriptions from multiple digital photographs and for automatically performing geometric character recognition on the impressions in order to derive meaning. The proposed system (more accurate, faster, and a more effective detective tool than any current alternative) comprises: (1) reconstructing detailed models of cuneiform tablets from multiple photographs; (2) isolating individual cuneiform strokes and distinguishing meaningful characters from background cracks; (3) classifying, against lexicons, the characters based on geometric characteristics; and (4) performing preliminary automatic word identification leading to translation of the texts. Ancient written records (Mesopotamian cuneiform tablets; Egyptian hieroglyphic carvings; or Roman political inscriptions) hold the keys to understanding our cultural heritage, but go unread because there are too many of them, their translation takes too long, and there are too few linguistic experts. Other arenas have objects with surface impressions (tire tracks, bootprints, and fossil remains) that must be measured for analysis and compared to like objects for identification, but such results cannot be efficiently obtained with current methods. The time lost, the inaccuracies created, and the potentially false interpretations presented can have dire consequences. The broader impact/commercial potential of this project are the significant benefits afforded to the target institutions allowing creation of accurate and precise 3D digital models of inscribed objects in their collections without the need for special equipment or experts. Since camera calibration, tripods, or special lighting are unnecessary, the process can be completed quickly, easily, and cheaply. Beyond the insight into the past that will inevitably accrue by using the proposed system, economic and technological advantages will motivate its commercial adoption. Museums, archaeological sites, and other collections need no longer pay for expensive scanners and their maintenance and upgrades, nor hire specialists to run the equipment and massage the results. Scholars need not travel to see inscriptions firsthand, still the traditional method of studying ancient written documents (photographs or drawings are insufficient for understanding the texts). The proposed system thus accomplishes many tasks the tools for which are unavailable; and it goes far beyond what has been attempted to aid the study and dissemination of information about ancient texts (e.g., Cuneiform Digital Library Initiative; Digital Hammurabi Project; Persepolis Foundation Archive). The software has extensive applications, such as in geology, forensics, and crime investigation, which we will explore in Phase II

The innovation, CUNAT (CUNeiform Automated Translator) attacks significant problems in a discipline wedded to conservative manual methods. Ancient Mesopotamian legal treatises, Egyptian medical records, and Roman political accounts--telling us exactly what happened in the words of the very people whose actions formed our history - languish untranslated, because there are too many of them, translation takes too long, and there are too few linguistic experts. "Cuneiform text genres include everything...from mathematical and grammatical exercises, beer recipes, international treaties, musical scores, legal codes, religious rituals, sales receipts, and astronomical tables....[M]useums...have acquired approximately 400,000 tablets, with thousands being unearthed every year.... [S]cholars continue to make unique and valuable contributions to the study of history, law, religion, linguistics, mathematics, and science" (Digital Hammurabi Project). CUNAT will allow anyone to create 3D models of artifacts and extract meaning from their impressions more effectively and quickly than existing tools. Phase II innovations include: extending the limits of image-based feature detection through unique algorithms based on 3D models automatically generated from photographs; isolating meaningful characters from noise across an array of surface shapes; classifying the characters of many dialects based on geometric characteristics; and performing automatic translation using graph-based algorithms and rules engines rather than traditional word-by-word methods.The broader/commercial impact of our innovation are new learning and analysis methods: a high-school student enters an art museum and sees an object with triangular-shaped surface incisions; the label reads "cuneiform tablet, 864BCE, Nimrud;" curious, the student aims a smartphone camera at the object and clicks; the images are uploaded into the project?s software app, which generates an interactive 3D model and an English translation of the inscription. The student may become the first person to have read the text and discover previously unknown astronomical knowledge detailed by the ancient scribe. Their software gives this research power and excitement of discovery to everyone, every institution, anywhere, anytime, because neither expensive special equipment nor complex calibration or lighting are necessary to produce results with unprecedented efficiency and accuracy. The innovations will address critical needs across the field of archaeology and produce a commercially viable product with applications to problems in museum exhibition, numismatics, and digital art history. In addition, accurate processes will be widely applicable, including forensics, paleontology, auto mechanics, personalized 3D printing, and carpentry. Beyond the insight into the past that will inevitably accrue, economic and technological advantages will motivate commercial adoption within and beyond our target market.