International contest seeks to unlock secrets of Herculaneum scrolls

Located east of Napoli, Mount Vesuvius was a massive catastrophe when it erupted in AD 79. It destroyed Pompeii and the adjoining city of Herculaneum. After claiming the lives of thousands, hundreds of ancient scrolls were carbonized in the library of a huge luxury mansion by the tremendous explosion of hot gas.

After proving that an artificial intelligence algorithm can extract letters and symbols from high-resolution X-ray photographs of the delicate, unrolled sheets, researchers are announcing a global competition to decipher the burned papyri, The Guardian reports.

Led by computer scientist Prof. Brent Seales of the University of Kentucky, researchers could read the ink on the surface and hidden layers of scrolls by training a machine-learning algorithm to spot subtle differences in the papyrus structure captured by the X-ray images.

“We’ve shown how to read the ink of Herculaneum. That gives us the opportunity to reveal 50, 70, maybe 80 percent of the entire collection,” said Seales. “We’ve built the boat. Now we want everybody to get on and sail it with us.”

Encouragement for solving the papyri

Seales' team is making its software and thousands of 3D X-ray scans of two rolled-up scrolls and three papyrus fragments available for the Vesuvius challenge. The prize money of $250,000 is intended to entice international research teams that can advance artificial intelligence and quicken the decoding of the only complete library that has survived from antiquity.

“We’re having a competition so we can scale up our ability to extract more and more of the text,” Seales said. “The competitors will be standing on our shoulders with all of our work in hand.”

Teams can compete for a $150,000 grand prize that will be given to the team that reads four chapters from the scrolls' innermost levels first before the end of 2023. For correctly identifying ink on the papyri from the 3D X-ray scans, progress awards include $50,000.

The burned papyri cannot be read because of the black ink used to create the scrolls, but infrared pictures of surface fragments have shown Greek letters and symbols. Seales's team trained their algorithm to recognize the letters from X-ray images alone using these data and X-ray images of identical fragments. After being taught, the system could detect new text concealed within the tightly bound scrolls.

“A human cannot pick this out with their eye,” Seales said. “The ink fills in the gaps that otherwise create a waffle-like pattern of the papyrus fibres. That pattern gets coated and filled in and I think that subtle change is what’s being learned.”