Digital humanities scholars have been using artificial intelligence for some time, but it’s been tough to unleash such tools at scale to probe ancient texts, particularly those in Hebrew and Aramaic, according to Michael Satlow, professor of Judaic studies and religious studies at Brown University.

Satlow, who self-identifies in a bio as an “amateur painter, gardener, cheesemaker and fisherman,” won a $249,956 grant from the National Endowment for the Humanities and the U.S.-Israel Binational Science Foundation to fund a three-year project on “knowledge transmission and cultural interactions through the ages,” an analysis of Jewish texts using AI.

“It is, in general, a problem using digital tools to analyze humanistic works because they are, by and large, unstructured,” Satlow told JNS. “Computers are very good at pattern detection, but their ability to find patterns in streams of words that may be in complex and unanticipated formats, and include other languages and references, is mixed.”

Difficulties deepen when one is using languages that have multiple dialects, conventions and abbreviations, like Hebrew and Aramaic, according to the professor.

“We could never have even contemplated a project like this, on this scale, without the development of large language models,” he told JNS.

He and colleagues aim to use the technology to look at more than 130,000 Hebrew and Aramaic texts dating between 200 and 2000 C.E.

The texts reflect centuries of debate across the Jewish world, so the researchers are also testing how artificial intelligence can compare different assumptions, arguments and even spelling variations.

“This is the advantage of using large language models,” Satlow said. “We will probably have to supply a number of examples to teach the large language model, but they are often quick learners.”

A central focus will be on extracting and analyzing citations, a defining feature of rabbinic writing.

“Rabbinic literature is highly referential,” the professor told JNS. “Rabbis cite other rabbis all the time, and later texts cite opinions of earlier rabbis.”

By determining who cites whom, and how those citations function within arguments, the research team hopes to identify key figures, places and periods in the transmission of knowledge.

Satlow is most excited about giving visual form to that network of citations. “Once we are able to describe that network, we can analyze it again using machine learning tools,” he told JNS.

He plans to work with Binyamin Katzoff, associate professor of Jewish studies at Bar-Ilan University; Maayan Zhitomirsky-Geffet, professor of information science at Bar-Ilan University; and Jonathan Schler, a professor of computer science at the Holon Institute of Technology. 

Most of the texts the four intend to analyze in the project will come from the already digitized Bar-Ilan Responsa database.

The post Brown professor wins federal grant to use AI to study Jewish texts appeared first on JNS.org.