AI Provides Insight Into Dead Sea Scrolls

Artificial intelligence (AI) is the key to unlocking many secrets of our past. This has been recently demonstrated when researchers at the University of Groningen cracked the code of the Dead Sea Scrolls with the help of AI. 

The Dead Sea Scrolls were discovered about seventy years ago, and they contain the oldest manuscripts of the Old Testament and other unknown Jewish texts. However, these scrolls have long been a mystery as scientists struggled to identify any of the authors of the scribes, which are unsigned. 

Mladen Popović is professor of Hebrew Bible and Ancient Judaism at the Faculty of Theology and Religious Studies at the University of Groningen. He is also director of the Qumran Institute at the University.

“They would try to find a “smoking gun” in the handwriting, for example, a very specific trait in a letter which would identify a scribe,” Popović says.

The findings of the research were presented in PLOS ONE on April 21. 

Cracking the Code

Popović created the project The Hands that Wrote the Bible, and along with colleague Lambert Schomaker, professor of Computer Science and Artificial Intelligence at the Faculty of Science and Engineering, and funding from the European Research Council, they set out to crack the code. 

Schomaker’s previous work involved developing techniques to enable computers to read handwriting from historical materials, a well as biomechanical traits that affect handwriting. 

PhD candidate Maruf Dhali was also involved in the research, and the team looked at one specific scroll called the Great Isaiah Scroll from Qumran Cave 1. 

“This scroll contains the letter aleph, or “a”, at least five thousand times. It is impossible to compare them all just by eye,” Schomaker says.

Because of this, the task makes much more sense for computers, which can analyze massive datasets. Digital imaging enables the microlevel of characters, like measuring curvature.

“The human eye is amazing and presumably takes these levels into account too. This allows experts to “see” the hands of different authors, but that decision is often not reached by a transparent process,” Popović says. “Furthermore, it is virtually impossible for these experts to process the large amounts of data the scrolls provide.” 

Training the Algorithm 

To train the algorithm, the team first separated the text from the background by using a state-of-the-art artificial neural network. It is able to keep the ink traces made by the scribe intact.

“This is important because the ancient ink traces relate directly to a person’s muscle movement and are person-specific,” Schomaker says.

The first analytical test demonstrated a possibility of more than one writer, so the data was then sent to Schomaker to recompute the similarities by using the patterns of letter fragments. The second process did in fact confirm two different variations. 

“When we added extra noise to the data, the result didn’t change. We also succeeded in demonstrating that the second scribe shows more variation within his writing than the first, although their writing is very similar,” Schomaker explained. 

The next step was to produce a visual analysis by creating “heat maps,” which incorporated the different variants of a character across the scroll. An average version of the character was produced for the first 27 columns and the last 27, and the human eye could spot a difference. This last sep brought together computerized and statistical analysis and human interpretation. 

An illustration of how heatmaps of normalized average character-shapes are generated for individual letters. Credit: Maruf A. Dhali

Previous scholars have believed that a new scribe started after the 27th column, but it had never been confirmed. 

“Now, we can confirm this with a quantitative analysis of the handwriting as well as with robust statistical analyses,” Popović said. “Instead of basing judgment on more-or-less impressionistic…