Artificial intelligence decodes 4,000-year-old cuneiform tablets

An artificial intelligence program time has been able to predict the missing more words and sentences from 4,500-year-old cuneiform tablets with amazing accuracy.

A utility and not a substitute for human experts

An artificial intelligence program has been able to predict the missing words, phrases and sentences from 4,500-year-old cuneiform tablets with amazing accuracy.

The archaeological tablets contain information about Mesopotamia dating back to between 2500 BC and 100 AD, and weather factors, wars and thefts caused the loss of many of these clay tablets, which hindered the ability of scholars to discover the secrets of the ancient civilization, due to the lost texts.

The artificial intelligence, which was taught how to read 104 languages, was provided with copies of ten thousand cuneiform tablets.

The program was able to accurately predict the missing words, phrases and sentences, similar to the way the phone suggests when adding the auto-complete feature.

Mesopotamia is one of the oldest known civilizations in the world and has given rise to the Sumerian, Assyrian and Babylonian empires. It was the birthplace of mathematics, astronomy, agriculture, written history, and many other disciplines.

Many Mesopotamian civilizations, including the Babylonians and Assyrians, spoke the Akkadian language, the oldest known Semitic language. They wrote in cuneiform, a form of writing that uses wedge-shaped characters, evidence of which can still be found today in the form of inscriptions on clay tablets.

In a paper presented in November at the Conference on Experimental Methods in Natural Language Processing, the authors stated, "These tablets are the master record from Mesopotamian cultures, including religious texts, bureaucratic records, royal decrees and more, so it is essential to understand these the language".

Over the millennia, some clay tablets were damaged, causing researchers to rely on contextual cues to manually fill in the missing text, a process the authors say is "subjective and time-consuming".

That's why researchers develop a deep-learning artificial intelligence system that can make informed guesses for missing words and phrases.

According to Britain's Daily Mail, Gabriel Stanovsky, a computer scientist at the Hebrew University of Jerusalem, told New Scientist that the AI   program is a "help tool" and not a substitute for human experts.

The researchers used a model that had already been trained in other Semitic well languages, such as Arabic, Hebrew, which share many similarities with Akkadian.

They first tested the system by masking existing sections of antique tablets, and the model completed the hidden sections with an accuracy of 89 percent.

The system was able to suggest contextually accurate words and phrases more to fill in the gaps or gaps.

Researchers have already used artificial intelligence to decipher the damaged inscriptions from ancient Greece, as scientists at the British company "DeepMind", a subsidiary of Google, trained a neural network called "Pythia" to guess the missing more words or letters from tens of thousands of Greek inscriptions that are between 1,500 years old. and 2,600 years.

The AI   achieved an accuracy of nearly 70 percent, compared to a group of Oxford-graduated researchers, who were right 43 percent of the time.