Apollo AI Model Helps Scholars Recover Lost Ancient Greek Texts

Libraries worldwide hold vast collections of fragmentary Ancient Greek papyri—many so deteriorated that their original content seems beyond recovery. Restoring them has always meant painstaking detective work by a small cadre of elite scholars. Now a new AI tool aims to transform that process.

A Language Model Built for Ancient Texts#

The Austrian Academy of Science is rolling out Apollo, described as the first large language model specifically designed for Ancient Greek. Built in collaboration with French AI firm Mistral and Sail Reply, Apollo was trained on roughly 600 million words drawn from historical manuscripts, papyri, and inscriptions spanning different periods and dialects.

The model will be accessible to academics through a free chatbot interface, enabling researchers to identify relevant document fragments and spot new research directions more efficiently.

Filling in the Gaps#

When papyri are damaged or faded, words disappear—and Ancient Greek texts lack word spacing, making restoration even trickier. Traditionally, scholars had to identify word boundaries, date the manuscript, understand its historical context, and manually select plausible replacements from reference materials. Apollo automates much of this reasoning: it can recognize stylistic signatures (Homeric verse versus Doric dialect, for instance) and suggest the statistically most likely words or phrases to complete gaps.

"The crucial point is that human competence needs to remain," says Anna Dolganov, a historian and papyrologist at the Austrian Academy of Science. Rather than imposing an answer, Apollo presents multiple candidate words, leaving the final choice to the expert.

Impact on Ancient Knowledge#

Experts expect the model to accelerate routine reconstruction, freeing scholars to focus on why documents matter rather than what they say. Armand D'Angour, a classics professor at the University of Oxford (which houses the world's largest ancient papyrus collection), calls the prospect "very exciting."

However, Apollo is unlikely to revolutionize our understanding of antiquity. Most unrestored papyri are mundane—personal letters, contracts, bureaucratic records—rather than lost masterworks. "If you were a layperson, you might think suddenly we'll get a few new plays by Sophocles, but that's not going to happen," notes Stephen Colvin of University College London. Instead, the tool will likely fill in small details about daily life and verify longstanding historical theories.