Dead Sea Scrolls: Artificial Intelligence sheds new light on their authors

An artificial intelligence-based paleographic project carried out by scholars in the Netherlands found a way to replace the human eye of with artificial intelligence.

Adolfo Roitman, curator of the Dead sea Scrolls, points at the original Isaiah Scroll, one of the Dead Sea Scrolls, inside a secured climate-controlled room in the Shrine of the Book at the Israel Museum in Jerusalem September 26, 2011. Developed in partnership with Google, the Israel Museum on Mond (photo credit: REUTERS/BAZ RATNER)
Adolfo Roitman, curator of the Dead sea Scrolls, points at the original Isaiah Scroll, one of the Dead Sea Scrolls, inside a secured climate-controlled room in the Shrine of the Book at the Israel Museum in Jerusalem September 26, 2011. Developed in partnership with Google, the Israel Museum on Mond
(photo credit: REUTERS/BAZ RATNER)
Were generations of scribes training together some 2,000 years ago in the Judean desert? Were some of the manuscripts known as the Dead Sea Scrolls produced as a team effort by two or more scribes working side by side in Qumran? And how many authors are behind the corpus of artifacts whose unearthing is considered one of the most crucial archaeological discoveries of the 20th century?
An artificial intelligence-based paleographic project carried out by scholars at the University of Groningen in the Netherlands hopes to find answers to many of these questions and to shed unprecedented light on the communities behind the text.
The first findings of the project were published in the journal PLOS ONE on Wednesday, solving a decades long riddle: the iconic ‘Great Isaiah’ scroll was written by two scribes and not one.
The Dead Sea Scrolls are a corpus of some 25,000 fragments unearthed in caves by the Dead Sea in the 1940s and 1950s. The artifacts include some of the most ancient manuscripts of the Bible, other religious texts that were not accepted in the canon as well as non-religious writings.
Paleography is the discipline that studies ancient writing. In the case of the Dead Sea Scrolls, it has been crucial to extract information ranging from the dating of the manuscripts to whether fragments carrying parts of the same texts originally belonged to the same scroll or to different ones.
Two 12x12 Kohonen maps (blue colourmaps) of (full) character aleph and bet from the Dead Sea Scrolls collection. (Credit: MARUF A. DHALI)
Two 12x12 Kohonen maps (blue colourmaps) of (full) character aleph and bet from the Dead Sea Scrolls collection. (Credit: MARUF A. DHALI)
This project represents the first attempt to replace the human eye of paleographers with an artificial intelligence analysis, as Prof. Mladen Popović, the head of the Qumran Institute of the University of Groningen, explained to The Jerusalem Post.
“Finding out how many scribes worked on a manuscript or were involved in the writing of the scrolls in general might sound like a trivial thing, but it opens up a whole new way of thinking about the Dead Sea Scrolls, not only as one collection, created for one group, but as different collections for different people,” Popović said. “We are just at the beginning, but it’s allowing us to see connections between the texts in a completely new perspective.”
Popović, who authored the study together with artificial intelligence experts Maruf A. Dhali and Lambert Schomaker, explained that they chose to start analyzing the Isaiah scroll both for its symbolic meaning  - 7 meter-long, it was one of the first seven scrolls found in 1947 and it is one of the best preserved – and for the fact that for decades scholars had been debating whether the artifact was produced by one or two scribes.
“This has been an undecided issue among scholars because the writing is so similar, but at the same time there are some differences in the ways the words are written in the two parts.” he explained. “Moreover, three lines at the bottom of column 27 were left blank and the new chapter, Isaiah 34, starts at column 28. The column also marks the beginning of a new sheet sown to the previous one. Normally, a new chapter would start in the same column.”

Stay updated with the latest news!

Subscribe to The Jerusalem Post Newsletter


“It’s a fascinating problem for paleography,” he added. “We all know that when one writes, they never write their letters exactly the same but there are some variations. Each person’s variations are different, but sometimes scribes can write very much in the same style, making it difficult for the human eye to distinguish between them. So this was first a test case for us.”
An illustration of how heatmaps of normalized average character-shapes are generated for individual letters. (Credit: MARUF A. DHALI)
An illustration of how heatmaps of normalized average character-shapes are generated for individual letters. (Credit: MARUF A. DHALI)
The AI experts developed an algorithm to analyze the patterns of these variations and were able to establish that the manuscript was indeed written by two different people, with the transition happening between column 27 and 29.
Asked whether this means that the scribes worked side by side on different parts of the biblical book which were then sown together, Popović said that while there cannot be any certainty, it is a plausible scenario, adding that the fact that the scribes’ handwriting was so similar might have been a sign that they trained together.
In order to train the algorithm, the experts are using the digitized images of the manuscripts provided by the Israel Antiquities Authorities, the body which was placed in charge of them on behalf of the state of Israel.
The team led by Popović is already at work to shed more light about the authors of other manuscripts, whether different manuscripts were written by the same scribes and issues related to the dating.
“The human eye is amazing and can see things that computer cannot see, but we cannot always realize what we're seeing, let alone explain what we're seeing, while the computer can quantify and give us the data.” he concluded. “The interpretation this data is on us.”