College Student Uses Machine Learning to Decipher Word in Ancient Roman Scroll

pcmag.com:

The Vesuvius Challenge is an academic contest to see who can translate ancient Roman scrolls recovered from the lost cities of Herculaneum and Pompeii through the use of machine learning. Researchers have been itching to see what's inside the scrolls since their excavation in 1752, and we now have some answers.

College student and former SpaceX summer intern Luke Farritor determined that one of the scrolls says the word Porphyras, which translates to purple. Youssef Nader, an Egyptian biorobotics grad student, verified the finding later, earning him second place.

Both students were inspired by work previously done by Casey Handmer, who was the first person to find evidence that ink exists on the scrolls.

For their efforts, Farritor and Nader were awarded the First Letters Prize, which carries with it a $50,000 reward for finding 10 legible letters in a 4-square-centimeter area on any scroll. Luke will get $40,000 and Youssef gets $10,000.

This is exciting for several reasons. The scrolls were buried and carbonized by the Vesuvius eruption in 79 AD. Despite being excavated hundreds of years ago, few of the scrolls have ever been read because most of them crumble into dust when they're opened. A few attempts were made in the intervening years after the discovery, but no one has attempted to open a scroll since the 19th century. The contents of the remaining scrolls are a complete mystery.

Dr. Brent Seales and his team took high-resolution X-rays of the scrolls and used a virtual unwrapping technique to digitally flatten out the scrolls. There was only one problem. The ink used by ancient Romans doesn't show up on X-rays. Researchers correctly estimated that machine learning could discern the difference between ink and