Ancient Content
Apr 30, 2025
AI Tool Reproduces Ancient Cuneiform Characters with High Accuracy
ProtoSnap, developed by Cornell and Tel Aviv universities, aligns prototype signs to photographed clay tablets to decode thousands of years of Mesopotamian writing.
Cornell University researchers report that scholars can now use artificial intelligence to “identify and copy over cuneiform characters from photos of tablets,” greatly easing the reading of these intricate scripts.
The new method, called ProtoSnap, effectively “snaps” a skeletal template of a cuneiform sign onto the image of a tablet, aligning the prototype to the strokes actually impressed in the clay. By fitting each character’s prototype to its real-world variation, the system can produce an accurate copy of any sign and even reproduce entire tablets.
Cuneiform, like Egyptian hieroglyphs, is one of the oldest known writing systems and contains over 1,000 unique symbols. Its characters change shape dramatically across different eras, cultures and even individual scribes, so that “even the same character… looks different across time,” Cornell computer scientist Hadar Averbuch-Elor explains. This extreme variability has long made automated reading of cuneiform a very challenging problem.
The ProtoSnap technique addresses this by using a generative AI model known as a diffusion model. It compares each pixel of a photographed tablet character to a reference prototype sign, calculating deep-feature similarities. Once the correspondences are found, the AI aligns the prototype skeleton to the tablet’s marking and “snaps” it into place so that the template matches the actual strokes. In effect, the system corrects for differences in writing style or tablet wear by deforming the ideal prototype to fit the real inscription.
Crucially, the corrected (or “snapped”) character images can then train other AI tools. The researchers used these aligned signs to train optical-character-recognition models that turn tablet photos into machine-readable text. They found the models trained on ProtoSnap data performed much better than previous approaches at recognizing cuneiform signs, especially the rare ones or those with highly varied forms. In practical terms, this means the AI can read and copy symbols that earlier methods often missed.
This advance could save scholars enormous amounts of time. Traditionally, experts painstakingly hand-copy each cuneiform sign on a tablet; the AI method can automate that process, freeing specialists to focus on interpretation. It also enables large-scale comparisons of handwriting across time and place, something too laborious to do by hand. As Tel Aviv University archaeologist Yoram Cohen says, the goal is to “increase the ancient sources available to us by tenfold,” allowing big-data analysis of how ancient societies lived – from their religion and economy to their laws and social life.
The research was led by Hadar Averbuch-Elor of Cornell Tech and carried out jointly with colleagues at Tel Aviv University. Graduate student Rachel Mikulinsky, a co-first author, will present the work – titled “ProtoSnap: Prototype Alignment for Cuneiform Signs” – at the International Conference on Learning Representations (ICLR) in April. In all, roughly 500,000 cuneiform tablets are stored in museums worldwide, but only a small fraction have ever been translated and published. By giving AI a way to automatically interpret the vast trove of tablet images, the ProtoSnap method could unlock centuries of untapped knowledge about the ancient world.
https://x.com/HistContent/status/1917528469316968892
Owen Gregorian
AI translates 5,000-year-old cuneiform tablets into English | Alexandra Gerea, ZME Science
Cuneiform is one of the earliest writing systems in human history. Archaeologists have traced its beginnings to around 3400–3300 BC, more than 5,000 years ago. It also lasted for a remarkably long time: the last securely dated cuneiform text comes from 75 AD.
Researchers have found hundreds of thousands of texts written in cuneiform, many of them in the Sumerian and Akkadian languages. Now, they’ve also trained a neural network that can translate digitized Akkadian cuneiform into English.
An old, mysterious language
The Akkadian language is one of the earliest known Semitic languages, a family that includes modern languages such as Arabic and Hebrew. It was spoken in ancient Mesopotamia, primarily in the Akkadian Empire that was situated in the region that is today parts of Iraq and northeastern Syria. Akkadian is named after the ancient city of Akkad, one of the major centers of the Akkadian civilization.
Akkadian was used for a wide range of purposes, from administrative and legal documents to literature and science texts. It was written using cuneiform script on clay tablets, and its decipherment in the 19th century opened up a new window into the ancient world, providing scholars with valuable insights into the history, culture, and scientific achievements of the time.
Meanwhile, Sumerian is one of the world’s oldest known languages, and it has the distinction of being a language isolate, meaning it has no known relatives. It was spoken in ancient Sumer, a region located in the southern part of what is now modern-day Iraq. The Sumerians are credited with establishing one of the world’s earliest civilizations around 4500 BCE, and their society flourished until about 2000 BCE.
Both languages used the cuneiform writing system, as did several other languages. Cuneiform is therefore a script rather than a language in itself (it’s not exactly an alphabet, either). It was adapted to write at least 15 languages, including Akkadian, Sumerian, Hittite and Elamite.
Translating the ancients
The story of deciphering cuneiform starts with the so-called Behistun Inscription. Discovered in Iran and dating back to the time of King Darius I of Persia (550 BC), this multilingual inscription included three types of script: Old Persian, Elamite, and Akkadian cuneiform. Old Persian was deciphered first, providing clues for the other two.
By 1857, confidence in the decipherment was strong enough for the Royal Asiatic Society to send the same Assyrian inscription independently to four scholars as a test, to see if they produce similar or comparable translations; they did.
But this doesn’t mean they could translate everything perfectly. Scholars have continued refining individual readings and interpreting difficult texts ever since.
For some researchers, deciphering the writing system was only the beginning. Hundreds of thousands of cuneiform tablets have been excavated, and many remain unpublished or untranslated. Researchers therefore turned to artificial intelligence to determine whether parts of this laborious process could be automated.
Cuneiform, meet AI
In recent years, language translations have come a long way — and AI is greatly accelerating these trends in automation. AI translations are nearing a watershed moment, with some pretty striking achievements. In the new study, Shai Gordin and colleagues from Ariel University described an AI model that can automatically translate Akkadian text written in cuneiform into English. For now, this is only available for this particular language (not all languages that use the cuneiform script work at the moment), but it’s still remarkable.
This was a follow-up to a previous study by Gordin and colleagues that also looked at how AI can be used to translate cuneiform. This time, two versions of the model were trained.
The first translated Akkadian that had already been transliterated into the Latin alphabet. Transliteration is a scholarly representation of the signs and their likely readings. The second model translated directly from Unicode cuneiform characters into English.
The first version gave better results in the study, achieving a score of 37.47 in the Best Bilingual Evaluation Understudy 4 (BLEU4).
The Bilingual Evaluation Understudy (BLEU) score is a metric used to evaluate the quality of machine-generated translations. It measures how closely a machine translation of a text matches a set of human-created reference translations. The score ranges from 0 to 1 (or 0 to 100), with higher scores indicating better translations. Even experienced human translators don’t usually get 100, and for a language such as cuneiform, 37 is good enough to get a decent translation.
The model achieves the best results in short and medium-length sentences. As the sentences get longer, the model struggles to grasp the entire context — although this can be trained in the future, researchers say. Another shortcoming is that the model also “hallucinates” — it creates outcomes that are syntactically correct but completely decoupled from the meaning of the original text. This is something that other engines, notably ChatGPT, also do sometimes.
Consider the following example:
Sentence 2,753
Source: UD 21-KAM2 LUGAL ina E2-DINGIR E2-DINGIR la ur-rad
Human translation: “On the 21st day the king does not go down to the House of God.”
Machine translation: “On the 21st day the king goes down to the House of God.”
In this case, the AI did a great job of translating most of the content. However, an error that likely occurred when cleaning the data for training caused the AI to miss the negation, completely altering the meaning of the sentence.
In the majority of cases, however, the translation was very useful as a first-pass of the text (though not as an independent translator). Researchers say the AI can be used by scholars or even by students who want to study this language in more detail.
What has changed since 2023
Since the original article was published in 2023, researchers have expanded AI’s role beyond translation. New systems are being developed to recognize signs, reconstruct damaged passages, organize digital collections and identify tablet fragments that may belong to the same work. This type of system has been used several times, with remarkable results.
In 2025, researchers Anmar Fadhil and Enrique Jiménez published a previously unknown hymn praising Marduk, Babylon and the Babylonian people. The work survived across 20 manuscripts written between the seventh and the second or first centuries BC. The platform helped the researchers locate and assemble related fragments, allowing them to recover about two-thirds of a poem that may originally have contained 250 lines. AI helped find the pieces; human archaeologists reconstructed, interpreted and translated the text.
Work has also begun to move beyond Akkadian. In 2024, researchers released SumTablets, a dataset pairing Unicode cuneiform signs with scholarly transliterations for 91,606 Sumerian tablets. Such datasets could support tools that suggest possible transliterations for Sumerian, although they still begin with signs that have already been digitally identified.
A 2025 research challenge called EvaCun tested language models on Akkadian and Sumerian tasks such as identifying dictionary forms and predicting missing portions of damaged texts. These are less eye-catching than complete English translation, but they address some of the repetitive work scholars perform when reconstructing fragmentary inscriptions.
In July 2026, another project called TabletCraft presented a bidirectional translation model trained on 116,000 Akkadian-English examples. Unlike the 2023 system, it can also attempt to translate English into Akkadian and render the result as cuneiform. The project is aimed partly at education and public engagement.
The long-term goal is to connect these stages: photograph a tablet, detect its wedges, identify the signs and language, reconstruct damaged passages and produce a translation that clearly marks uncertainty. No current system can complete that entire process reliably without substantial human supervision.
Still, the progress since 2023 shows that AI can do more than generate a rough English translation. It can help scholars search enormous collections, identify possible connections between fragments and process texts that might otherwise remain unread for decades.
Read more:
https://zmescience.com/science/news-...ranslation-rep
[url]https://x.com/OwenGregorian/status/2086073535613542903
Re: A.I. is Progressing Faster Than You Think!
Reply With Quote
).
• They're Banning THESE Home Servers
Re: A.I. is Progressing Faster Than You Think!
Bookmarks