New Israeli Research Can Decode and Write in Cuneiform

In modern times, spelling can vary quite a bit from place to place. A word, for example, may be spelled color in the US, but colour in the UK, for example. The ancient world also didn’t produce written language in a standardized way, according to Discover Magazine. Cuneiform, one of the oldest writing systems, consists of more than 1,000 characters that vary widely by location, age, and culture.

Maybe this local variability was just as intelligible to the ancient world as British English is to Americans. But it’s not necessarily intuitive to modern scholars trying to decode hundreds of thousands of old Mesopotamiam documents. Now, an AI translation and transcription tool developed by researchers at Tel Aviv University and Cornell University in Ithaca, New York, makes it possible for archivists to access ancient knowledge simply by taking a picture.

What is Cuneiform
Cuneiform is a writing system from Mesopotamia that rivals hieroglyphs in its antiquity. Ancient documents penned in cuneiform include everything from mundane business records to the first known work of literature, the Epic of Gilgamesh. There are thousands of unique cuneiform characters making translation both time-consuming and labor intensive.  

Hadar Averbuch-Elor, a professor at Cornell Tech, explained in the Cornell Chronicle, that, “When you go back to the ancient world, there’s a huge variability in the character forms. Even with the same character, the appearance changes across time, and so it’s a very challenging problem to be able to automatically decipher what the character actually means.”

ProtoSnap 
To mitigate these challenges Averbuch-Elor’s team partnered with a team from TAU to create an AI tool they called ProtoSnap. This builds on previous research from TAU that created an AI model that translated Akkadian text written in cuneiform into English.While this machine learning tool was helpful, it was still not enough to translate all of the ancient documents that have been found. That’s where ProtoSnap comes in.

ProtoSnap uses computer vision to scan the pixels in a picture of a cuneiform table and compares them to a series of prototype cuneiform characters, according to Phys.org. Then, it transcribes and translates the original characters into standardized, recognizable characters, projecting the transcription right on top of the original writing.

The original and translated scripts are also then used to train future AI models. As a result of this ongoing training, ProtoSnap has been able to translate even rare and localized cuneiform characters. 

Rediscovering Ancient Insights
By instantly standardizing cuneiform text and making it legible, ProtoSnap facilitates researchers to quickly and efficiently decode ancient texts. This is important because there are up to 500,000 cuneiform tablets that have been discovered but are yet to be read or translated. Some may contain ancient wisdom and legends as historic as the Epic of Gilgamesh was.

Yoram Cohen, an archaeology professor at Tel Aviv University who worked on the project explained in the Cornell Chronicle that, “At the base of our research is the aim to increase the ancient sources available to us by tenfold. This will allow us, for the first time, the manipulation of big data, leading to new measurable insights about ancient societies — their religion, economy, social and legal life.”

Who knows what ancient insights are hidden in the 500,000 cuneiform documents written thousands of years ago? Now, by leveraging the power of modern technology to decode ancient script, researchers have a unique chance to reveal these bygone secrets, hidden from view for millennia.