Google Unveils AI Tool to Restore Lost Words in Ancient Roman Inscriptions

A New Tool for Deciphering Ancient Rome

Artificial intelligence is now being used to unlock the secrets of ancient Rome. A new AI tool developed by Google DeepMind, called Aeneas, helps scholars fill in missing words from ancient inscriptions and estimate historical data about these texts. The name Aeneas comes from a hero in Roman mythology who defended his city during the Trojan War.

Ancient inscriptions are crucial for understanding the Roman world, but they can be challenging to decipher. Historians often rely on “parallels,” which are other texts that share similar wording, syntax, or origin. Aeneas was created to speed up this process. According to a statement from Google DeepMind, the tool can "reason across thousands of Latin inscriptions, retrieving textual and contextual parallels in seconds."

Each year, around 1,500 ancient Roman inscriptions are discovered, as noted in a paper published this week in the journal Nature. These inscriptions are unique because they were written by people from all social classes, not just the victors of history.

To help historians interpret these texts, Aeneas is trained on some of the largest databases of Latin epigraphy, such as the Epigraphic Database Roma, Epigraphic Database Heidelberg, and the Epigraphik-Datenbank Clauss-Slaby. Google DeepMind’s team, led by AI researcher Yannis Assael, compiled these records into a single dataset featuring over 176,000 Latin inscriptions from across the Roman world. When given a new inscription, the model retrieves a list of parallels based on text and provenance, then uses those to guess what might be missing.

Researchers found that Aeneas can restore damaged inscriptions with 73 percent accuracy when up to ten Latin characters are missing. If the gap length is unknown, the accuracy drops to 58 percent. The model can also attribute an inscription to one of 62 ancient Roman provinces with 72 percent accuracy and estimate the date of a text within 13 years.

In one test, Aeneas analyzed the Res Gestae Divi Augusti, a first-person account by Emperor Augustus. The exact date of the inscription is still debated, but Aeneas provided a distribution of possible dates with two peaks: one between 10 and 20 C.E., and another between 10 and 1 B.C.E. These ranges align with existing scholarly hypotheses.

“This was a jaw-dropping moment for us,” said Thea Sommerschield, a historian at the University of Nottingham who collaborated on Aeneas.

Mary Beard, a Cambridge University historian, described the potential of Aeneas as “transformative” for scholars. She noted that breakthroughs in this field have traditionally relied on the memory and judgment of individual scholars, supported by traditional databases. Aeneas, she said, opens up entirely new horizons.

Jonathan Prag, an ancient historian at the University of Oxford, believes Aeneas could allow more people to work on existing inscriptions. However, he cautions that the tool should be used carefully. “The only way you can do it without a tool like this is by building up an enormous personal knowledge or having access to an enormous library,” Prag told the Guardian. “But you do need to be able to use it critically.”

Aeneas is built on a project called Ithaca, a Google initiative that used AI to interpret Greek inscriptions. An interactive version of Aeneas is available for free at predictingthepast.com, and the code and dataset have been open-sourced. This makes the tool accessible to researchers and enthusiasts alike, offering a powerful new approach to studying ancient texts.