Machine Learning Can Find Gaps in Ancient Greek Texts

Machine learning (ML) could bring lost texts to life, from imperial decrees to poems by Sappho.

At certain times in history, humans began to use written texts to keep accounts.

These accounts can give modern scholars clues as to how people in ancient societies conducted their days. But that’s only if the artifacts can be deciphered. Many have been weather eroded or broken and parts are missing. Modern scholars use a variety of tools to determine the content of the original text. It is almost always a long and tedious process.

Machine learning has the potential to contribute to the toolbox of historians involved in analyzing ancient texts and using them to understand processes such as the development of imperialism or the nature of cult activity.

Researchers have developed a powerful machine learning-based system that can fill in gaps in ancient Greek inscriptions and determine when and where they came from.

Machine learning is a powerful technology that can help historians, researchers, and scientists ask questions about human history and make comparisons with existing evidence.

DeepMind researchers working alongside scientists from the University of Venice, the University of Oxford and the Athens University of Economics and Business, have developed a machine learning application that can help historians to fill in the gaps of missing text in stone, metal or pottery artifacts.

Most of the surviving inscriptions have been damaged over the centuries, but machine learning can now restore the missing parts.

Using machine learning, the researchers achieved 62% accuracy when used alone to fill in gaps in the inscriptions, and 72% accuracy when the system’s suggestions were interpreted by a historian.

The researchers trained the machine learning algorithms using 60,000 Greek texts from 700 BC to 500 AD. Machine learning has helped researchers assign a text to a time and place using clues found in the text and from other sources.

It is also possible to develop machine learning systems that could provide information on the authorship of texts.

Using machine learning will help us sift through our vast digitized record of our history and fill in all those blanks, connecting events from centuries ago as richly as those from just minutes ago.

Instead of replacing humans, artificial intelligence will help us better understand ourselves, how we got here, and what it means to be human.

Sherry J. Basler