Detail of a marble stele inscribed with a decree of the Athenian boulē, c. 440–425 BCE, in the Athens Epigraphic Museum (via Wikimedia Commons)

A new deep neural network could help historians restore ancient inscriptions, identify their original location, and date them with impressive accuracy. Named Ithaca and developed by DeepMind, a British artificial intelligence (AI) research lab, and researchers at Ca’ Foscari University of Venice, Oxford University, and Athens University, the breakthrough demonstrates how machine learning can be deployed synergistically to aid historians’ work.

In a paper published in the journal Nature earlier this month, the researchers found that Ithaca can restore damaged texts with 62% accuracy, identify their place of origin with 71% accuracy, and date them within 30 years of their estimated ranges. But when used in tandem with a historian’s expertise, the accuracy of the combined effort jumped significantly.

“Ancient history relies on disciplines such as epigraphy — the study of inscribed texts known as inscriptions — for evidence of the thought, language, society and history of past civilizations,” the authors of the paper write. “However, over the centuries, many inscriptions have been damaged to the point of illegibility, transported far from their original location and their date of writing is steeped in uncertainty.”

The algorithm was trained on “the largest digital dataset of Greek inscriptions” from the Packard Humanities Institute, according to a press release, which allows it to overcome individual biases and prior errors while relying on decades of past research.

The tool promises to help epigraphers in their efforts to reconstruct fragmentary texts that have survived over the centuries and millennia. Also crucial to epigraphers’ work is dating an inscription (chronological attribution) and identifying its place of origin (geographic attribution). Conventional epigraphic methods are often very elaborate and tedious, meaning that a tool like Ithaca can do much to streamline the process.

This inscription contains a decree about the Acropolis of Athens and dates from 485/4 BCE. (courtesy DeepMind)

One key use case for Ithaca, therefore, is more precisely dating texts that as of now can only be ascribed wide date intervals. To prove how Ithaca could meaningfully contribute to debates surrounding dating, the researchers used the neural network on a set of Ancient Greek decrees whose date is significant to our understanding of classical Athenian political history. There is widespread disagreement among historians over whether those inscriptions were made before or after 446/445 BCE. The researchers showed that their estimates aligned with “the most recent dating breakthroughs,” and were more precise than competing dating methods.

A free interactive version of Ithaca has been made available in partnership with Google Cloud and Google Arts & Culture, where users can input Ancient Greek text with missing characters and command the tool to date, locate, and restore it. Researchers also open sourced the code to facilitate additional research.

“We believe machine learning could support historians to expand and deepen our understanding of ancient history, just as microscopes and telescopes have extended the realm of science,” Yannis Assael, staff research scientist at DeepMind, said in a statement. “Ancient Greece plays an instrumental role in our understanding of the Mediterranean world, but it’s still only one part of a vast global picture of civilisations that could be explored.”

The Latest

Stories That Need to Be Told

Stories That Need to Be Told

Artist Minouk Lim wants to offer a very different perspective on how one might deal with a grim history whose effects continue to be felt in the present.

Required Reading

Required Reading

This week: Should Washington have a national memorial for gun violence? Have cats used us to take over the world? What is Cluttercore? And more.

Jasmine Liu

Jasmine Liu is a staff writer for Hyperallergic. Originally from the San Francisco Bay Area, she studied anthropology and mathematics at Stanford University. Find her on 

Leave a comment

Your email address will not be published.