An ancient Middle Eastern text that is over 5,000 years old may finally be decoded, thanks to the implementation of artificial intelligence (AI). Experts estimate that there are around a million cuneiform tablets worldwide, left behind by ancient Mesopotamians. However, the task of translating and cataloging these tablets is a laborious process for archaeologists. It is estimated that up to 90 percent of cuneiform texts remain untranslated to this day.
Researchers in Germany have developed a new approach using AI that enables computers to recognize cuneiform characters and make the contents of these ancient tablets searchable, resembling a website. This breakthrough could potentially lead to the digitization and assembly of larger libraries of ancient texts, allowing for the discovery of previously unknown details about ancient life. The tablets contain a wide range of information, from significant events such as temple construction to seemingly trivial matters like customer service complaints.
The German team trained an AI program on two cuneiform languages: Sumerian and Akkadian. Both languages were used in writing throughout Mesopotamia, a region that comprised modern-day Iraq, parts of Iran, Kuwait, Syria, and Turkey. The cuneiform tablets are written in multiple languages and are thousands of years old, presenting a challenge for researchers who aim to decipher them.
The cuneiform characters were carved into three-dimensional clay tablets, which, combined with their age and weathering, makes scanning them into a computer a difficult task. The researchers utilized 3D models of approximately 2,000 tablets to train their AI program. Similar to using a smartphone camera to convert a handwritten note into a text document, the program scans the text on the tablets and transcribes it. The researchers did not focus on translating the tablets themselves but rather aimed to facilitate the work of other translators by creating an efficient tool.
Deciphering the text on the tablets can be particularly challenging due to their weathered clay surfaces and the lack of contextual information. Working with such tablets has been a cumbersome process until now, as researchers had to rely on knowing exactly what they were looking for and where to find it. The German team used an open-access set of 3D scans containing Sumerian cuneiform tablets, originating from southern Mesopotamia, now south-central Iraq. The newfound capabilities of the AI program will greatly aid humanities researchers in understanding the intricacies of life in ancient Mesopotamia.
One of the main challenges the researchers faced was training the AI to recognize the unique wedges and signs that form cuneiform characters. By inputting 21,000 signs and 4,700 wedges into the program, they created a comprehensive data set that can be utilized by future cuneiform scholars. Extensive testing on various tablets revealed that the AI program achieved an accuracy rate of around 76 percent when detecting cuneiform wedges and signs. It even exhibited reliable performance when working with lower-quality photographs, much to the researchers’ surprise.
Moving forward, the team plans to train the AI on a larger sample of tablets to enhance its accuracy. They believe that the relatively small number of tablets they used initially may have limited the program’s precision. Moreover, they could improve accuracy by segmenting the tablet images into smaller portions, reducing the amount of information the AI program needs to process at once.
The study, published in the November issue of Eurographics Workshop on Graphics and Cultural Heritage, marks a significant step forward in unraveling the mysteries of these ancient Middle Eastern texts. With AI technology aiding in the translation and searchability of cuneiform tablets, researchers will have access to an unparalleled wealth of information about early civilizations, allowing them to gain valuable insights into ancient societies.
The deciphering of this 5,000-year-old text through the use of AI technology represents a major breakthrough in the field of archaeology and linguistics. By employing innovative methods to train computers in recognizing cuneiform characters, researchers are now able to unlock the secrets contained within these ancient tablets. The digitalization of these texts opens up new avenues of research and promises to shed light on various aspects of ancient life. As this technology continues to advance, historians will be able to paint a more comprehensive picture of our ancient ancestors and their fascinating civilizations.
