Article contents:
- Introduction: A window to the past thanks to technology
- What is the Cairo Geniza and what is its historical importance?
- Challenges of studying ancient manuscripts
- The MiDRASH project: Artificial intelligence in the service of history
- How does AI transcription technology work?
- Potential discoveries and a “medieval Facebook”
- Frequently asked questions
Introduction: A window to the past thanks to technology
Researchers in Israel hope to uncover new discoveries about Jewish history by harnessing the power of artificial intelligence. A digital database of thousand-year-old manuscripts is now being uploaded to a new transcription tool that uses machine learning to read and decode ancient texts that were previously difficult, if not impossible, to analyze at scale.
At the heart of this effort is the “Cairo Geniza,” considered the world’s largest collection of medieval Jewish documents. This historical treasure trove is now accessible for deeper and faster study than ever before.
What is the Cairo Geniza and what is its historical importance?
A “geniza” is a repository in a synagogue designated for important documents that are ultimately meant to be buried. The geniza found in the Ben Ezra Synagogue in historic Cairo enjoyed a dry climate that was ideal for preserving old paper. Medieval Cairo was a global center of trade and science and home to a thriving Jewish community.
The Cairo Geniza contains a staggering collection of more than 400,000 documents, including civil records, religious debates, personal correspondence, and even some documents written by the great Jewish philosopher Moses Maimonides himself, who served as a physician to the family of Sultan Saladin.
Scholars discovered this treasure trove in the late 19th century, and it has been the subject of ongoing study ever since, but its sheer scale means significant gaps remain in our understanding of its contents.
Challenges of studying ancient manuscripts
Although the entire collection has already been digitized and is available online as images, studying it has been a painstaking and time-consuming process. Most of its items have not been cataloged, and many are unorganized fragments of longer documents. Only about one-tenth of the collection has been transcribed by hand by scholars over more than a century.
The documents are written in multiple languages—including Hebrew, Arabic, Aramaic, and Yiddish—and in a wide variety of different handwritings, making them difficult to read even for trained experts.
The MiDRASH project: Artificial intelligence in the service of history
This is where the MiDRASH transcription project comes in. Funded by the European Research Council, it brings together researchers from several universities and institutes. The project relies on the digital database of Cairo Geniza documents housed at the National Library of Israel.
“We are constantly trying to improve the machine’s ability to decode ancient texts,” says Daniel Stökl Ben Ezra of the École Pratique des Hautes Études in Paris, one of the project’s principal researchers. The project has already made significant progress and could open up the documents to many different researchers.
How does AI transcription technology work?
The system works by training an artificial intelligence model to recognize patterns in ancient texts and transcribe them. The model is fed digital images of manuscripts along with pre-existing manual transcripts to learn how to associate written shapes with their corresponding words.
The more data the model is trained on, the higher its accuracy becomes. Transcriptions of the most difficult manuscripts are reviewed by researchers to ensure accuracy, which in turn helps improve the AI’s training in a continuous feedback loop.
Once texts are transcribed, researchers can access the entire collection and analyze it much more quickly. They can search for specific names or words, cross-reference different documents, and piece fragments back into more complete documents.
Potential discoveries and a “medieval Facebook”
The potential for discovery is enormous. One document transcribed by the project is a 16th-century Yiddish letter from Rachel, a widow in Jerusalem, to her son in Egypt, with his response written in the margins recounting his efforts to survive a plague sweeping through Cairo. Such personal glimpses provide invaluable insights into daily life during that era.
“The potential for reconstruction, to create a kind of ‘medieval Facebook,’ is right before our eyes,” Stökl Ben Ezra says. By linking names and events across thousands of documents, researchers can reconstruct the social, economic, and religious networks of society at that time in unprecedented detail.
He added that “modern translation capabilities are now incredibly advanced, and connecting all of this becomes much more feasible and accessible to the casual reader, not just academics,” opening up this historical treasure to a broader audience.
Frequently asked questions
1. What is the Cairo Geniza?
It is the world’s largest collection of medieval Jewish documents, found in the Ben Ezra Synagogue in Cairo, containing over 400,000 documents spanning a thousand years of history.
2. Why is artificial intelligence being used to study it?
Due to the collection’s massive size and the variety of languages and handwritings, manual transcription is far too slow. AI drastically accelerates the transcription process, allowing researchers to analyze the entire collection.
3. What kind of information can be discovered?
Details about daily life, trade, social relations, religious debates, and even the biographical reconstruction of individuals who lived centuries ago can be discovered.
