The researchers describe it as a significant step toward preserving Mesopotamia’s cultural heritage.

Artificial intelligence has been used to accurately translate ancient cuneiform texts on stone tablets into English by Israeli researchers at Tel Aviv University and Ariel University. “Another major step toward the preservation and dissemination of the cultural heritage of ancient Mesopotamia,” they describe the undertaking.

In the May issue of PNAS Nexus, the researchers presented the first neural machine translation of Akkadian into English. Arkeonews noted that their outcomes are “equal to those produced by an average machine translation from one modern language to another.”

Over the most recent 200 years, archeologists have found a huge number of texts that tell the historical backdrop of old Mesopotamia, a large portion of them written in Sumerian or Akkadian, made sense of the writers. The vast majority, on the other hand, remain untranslated due to their fragmentary nature, a lack of skilled translators, and their sheer volume. Moreover, cuneiform signs are polyvalent, there are a wide range of sorts of messages, and, surprisingly, the names of individuals and spots can be composed as perplexing sentences.

In an email to Artnet News, one of the authors, Luis Sáenz, of the Digital Pasts Lab in the Department of Land of Israel Studies and Archaeology at Ariel University, stated, “First, let me state that we believe that A.I. will not replace philological work.” We wish to accelerate the procedure. In the future, we hope that artificial intelligence will be able to assist both Assyrians and non-Assyrians in reading cuneiform texts.

This is just the most recent instance of scientists working with the most recent materials and the most recent tools. College of Kentucky specialists fostered an A.I. framework to peruse look over that were burned when Mount Vesuvius ejected in the year 79, and archeologists in Italy are dealing with a robot that utilizes A.I. to remake antiquated relics from their dispersed shards.

Sáenz asserts, “Obviously, the model has limitations.” Ancient languages are difficult to translate because we only have fragments of texts and no context. A.I. will need more tools to digitize data published in papers in the future in order to keep training the model and improve results. Fragments with only one or two lines are extremely difficult to work with. Additionally, a public-friendly web-based platform is essential.