Vienna, Sept. 28 (SANA) Researchers have unveiled an artificial intelligence model that suggests missing words and passages in damaged ancient Greek texts, a tool they say could help scholars study historical documents that remain incomplete.
The Austrian Academy of Sciences said Apollo was developed with French AI company Mistral AI and technology firm SAIL Reply, in cooperation with David Smith of Northeastern University in the United States. The first version is available for researchers and the public.
Trained on about 600 million words of historical Greek, the model analyzes the surviving text and proposes possible restorations for scholars to review. It can also distinguish between different periods and styles of Greek writing, the academy said.
In a blind evaluation, experts rated Apollo’s suggestions at least as good as human reconstructions in 77% of cases involving previously unknown gaps, according to the academy. Its results are suggestions for expert assessment, not definitive readings of the original documents.
The researchers demonstrated the system on a Greek birth declaration from Roman Egypt dated to 149 CE, a philosophical text carbonized during the eruption of Mount Vesuvius and a damaged inscription from the Black Sea region concerning Roman law.
The academy said museums and libraries hold hundreds of thousands of ancient inscriptions and more than a million papyrus fragments, many of which are incomplete or have yet to be deciphered. It is developing further tools to help analyze previously unread papyri and search across Greek texts.
R.H/M.F
