Scientists Develop AI Tool That Can Identify And Reconstruct Ancient Texts

By Toby T

Researchers at Germany’s Ludwig Maximilian University have developed Fragmentarium – an artificial intelligence (AI) system that can break down the meanings behind several ancient texts. 

Source: @serinus/Pexels

The algorithm, which is currently making the rounds on the internet, has been in development since 2018. It was built as part of efforts to digitize all Babylonian cuneiform tablets, and it currently holds the potential to combine some of the oldest and most popular stories written – up to and including the Epic of Gilgamesh. 

As the researchers explained, the major issue with deciphering Babylonian texts is that the narratives are written on clay tablets. And today, these tablets are only available as fragments. Besides that, the texts themselves are compiled in two complex writing systems – Akkadian and Sumerian. So, the task of compiling them can be especially challenging. 

Hitherto, the texts had been deciphered by copying characters on paper, then comparing these transcripts with others to see the fragments that belonged together and find possible ways to fill content gaps. 

With Fragmentarium, the process is significantly easier. The algorithm has digitized 22,000 text fragments so far, sifting through images and properly assembling text fragments in seconds. A statement described the algorithm as groundbreaking, with researchers sharing their belief that it could help to preserve and refresh Babylonian literature. 

Source: @photography-maghradze-ph-1659410/Pexels

Fragmentarium was especially popular last November when it recognized a fragment that belonged to the Epic of Gilgamesh – the oldest surviving piece of literature on Earth. In another impressive feat, it helped to identify a hymn to the city of Babylon. 

Enrique Jiménez, one of the lead researchers in the project, explained that the algorithm has so far helped to find 15 new hymn fragments, cutting down a process that would have taken between 30 to 40 years into a few minutes.