Send friends 5,000-year-old emojis with new Google AI hieroglyphic app

Users can play with hieroglyphics, practice identifying and drawing them and even send their very own message in ancient Egyptian to friends.

JULY 26, 2020 17:58

A wall of a chamber of the tomb of Mehu is seen after it was opened for the public at Saqqara area near Egypt's Saqqara necropolis, in Giza, Egypt September 8, 2018 (photo credit: REUTERS/MOHAMED ABD EL GHANY) — A wall of a chamber of the tomb of Mehu is seen after it was opened for the public at Saqqara area near Egypt's Saqqara necropolis, in Giza, Egypt September 8, 2018

(photo credit: REUTERS/MOHAMED ABD EL GHANY)

Google Arts and Culture has launched an application aimed at employing artificial intelligence to help scholars translate Egyptian hieroglyphics and promoting knowledge about the topic to the general public.

The challenge of deciphering hieroglyphics has captured generations of scholars over the centuries. The breakthrough came when Napoleon discovered the Rosetta Stone, carrying the same text in Egyptian writing and ancient Greek, in 1799.

However, more than 200 years later, the work of translating 4,000 years of texts remains colossal. To this day, experts have to decipher and study each inscription manually.

Fabricius – the tool launched by Google in cooperation with the Australian Center for Egyptology at Macquarie University, digital-production company Psycle Interactive, video-game company Ubisoft and Egyptologists from all over the world – is designed to change that.

The application is addressed to both scholars and the general public.

For the former, the idea behind it is to apply machine-learning technologies to allow the algorithms to learn how to visually recognize the different symbols and produce a translation.

“Specifically, Google Cloud’s AutoML technology, AutoML Vision, was used to create a machine learning model that is able to make sense of what a hieroglyph is,” reads a Google press release. “In the past you would need a team of data scientists, a lot of code, and plenty of time, now AutoML Vision allows developers to easily train a machine to recognize all kinds of objects.”

Moreover, the software has been released as open source to encourage further developments.

“Digitizing textual material that was up until now only in handwritten books will completely revolutionize how Egyptologists do business,” Dr. Alex Woods of the Australian Centre for Egyptology told the BBC. “Digitized and annotated texts could potentially help us to reconstruct broken texts on the walls and even to discover texts we didn’t know were there.”

The first goal of the application is to be trained to extract hieroglyphics from images and create a facsimile of them. Researchers are therefore invited to upload the pictures in the system, including of very damaged symbols that cannot be identified manually.

Fabricius then works to compare the results with the hieroglyphics included in its library, while the user can correct the identification manually.

Finally, the system offers support to the Egyptologist in the process of translation by matching sequences and blocks of text to available dictionaries and published works.

However, the app is also designed for the general public. Besides offering stories and resources on ancient Egypt both for the occasional user and for educators and teachers, visitors to the website can play with the hieroglyphics, practice identifying and drawing them and even send their very own message in ancient Egyptians to their friends.

“We’re continuing to work on the tools and models as we gain more feedback and data,” the creators explained on the website. “But Egyptian hieroglyphs are just the first stage of Fabricius – we’re confident that the same approach can be applied to other ancient languages, too!”