Google Translate-like AI tool can interpret Cuneiform, a 5000-year-old writing system!

A recently developed AI tool might be able to unearth mysteries of the past by translating a 5000-year-old writing system into English!

Cuneiform can be found inscribed on clay tablets that are more than 3000 years old. (Gai Gutherz)

The artificial intelligence (AI) revolution has taken the world by storm. The emergence of AI tools such as ChatGPT and Google Bard has had a profound impact on how we access and process information, with everything now available with just a prompt and click. While these amazing tools can help us in accomplishing simple tasks in our daily lives, they can also help uncover the secrets of the past. A new and groundbreaking AI project, by a team of archaeologists and computer scientists from Israel, aims to do exactly that by translating a nearly 5000-year-old writing system into English in a manner of seconds, meaning it is a supercharged version of Google Translate!

Translating Akkadian Cuneiform

Cuneiform is a logo-syllabic script with characteristic wedge-shaped impressions that is over 5000 years old. It was in use from the bronze age to the common era, a period that last nearly 3000 years but hasn't been used in 2000 years. There are millions of clay tablets inscribed with cuneiform in museums, libraries, and galleries today. However, due to the extremely low number of Akkadian readers, translating them into understandable language today has been difficult, but not anymore.

The Google Translate-like tool was developed as a thesis project at Tel Aviv University. In a research published in the peer-reviewed PNAS Nexus, from the Oxford University Press, Gai Gutherz, a computer scientist who was part of the team that developed the program said, “What's so amazing about it is that I don't need to understand Akkadian at all to translate [a tablet] and get what's behind the cuneiform. I can just use the algorithm to understand and discover what the past has to say.”

Developing  the AI tool

According to the research, the AI tool uses a mathematical formula known as Neural Machine Translation (NMT), an artificial neural network that is used to translate text from one language to another. While it is difficult to judge what a good translation is, researchers used a machine translation evaluation tool called Best Bilingual Evaluation Understudy 4 (BLEU4) to determine the accuracy.

Accurate BLEU4 scores range from 0-100 with 0 being the lowest and 100 being perfect, which is unprecedented. The Akkadian translation tool returned a score of 36.52 for cuneiform to English, and a score of 37.47 for transliterated cuneiform to English, which is “fairly good”, according to Gutherz.

Hurdles faced

Since the language is so old, the biggest challenge was to find material to train the AI model on. Researchers combed through images of Cuneiform tablets and even drew samples from an online database from the University of Pennsylvania called Open Richly Annotated Cuneiform Corpus (ORACC). The AI model was trained on 50,544 sentences while 2,808 sentences each were reserved for validation and testing purposes.

Moreover, Akkadian was used for over 3000 years, which led to vast differences in the cuneiform symbols and even dialects during this period, which added to the complexity of translating such an old writing system.

Where to get it?

If you wish to try this groundbreaking AI tool, then an early demo version of it is available on The Babylon Engine. Moreover, if you wish to develop a tool like this of your own, then you can find its source code on GitHub on Akkademia and the Colaboratory.