The Future of the Talmud
A Digital Humanities Case Study
This is our Spring 2020 demo. Spring 2021 demo here!
What am I looking at?
You're looking at the current prototype of our digital implementation of the Talmud and its unique visual form. This is a native web implementation, meaning your browser recognizes it like any other text on the web. (Try copying and pasting, for example.) Aside from the rendering framework, we've built an algorithm to render almost any of the 2000+ pages of the talmud. We plan to release these in an open-source and extendible library, which will make it the first open-source native web implementation of the tzurat hadaf. Our demo also allows you to click on a word in the main text to see English translations for both sentences (⬄) and words (W).
What is the Talmud, anyway?
The Talmud is the core canon of Jewish law, lore, and wisdom. Compiled around 500CE, it's a massive collection of commentary on the Mishna (an earlier law code compiled around 200CE) and the Hebrew Bible. Aside from constituting the standard curriculum for Jewish learning and the launching point for Jewish legal discussions, it is also studied by a growing number of laymen and scholars alike in the ever-popular Daf Yomi (daily page) learning program. The vast majority of this study is done through the printed text, paginated into 2,711 double-sided pages. The page (called the daf) has a distinct and standardized visual form, called the tzurat hadaf, which visually relates commentaries, related texts, and references. Hover to learn about the standard daf: the main text • Rashi • Tosafot • pagination
Written primarily in Aramaic, the main text of the Talmud has voices from over 500 years interacting with one another: debating on Jewish law, telling stories, and interpreting the text of the Hebrew Bible and the Mishna.
Rabbi Shlomo Yitzchaki, known by the Hebrew acronym Rashi, lived in France in the 11th century. His commentary on the Talmud soon came to be a staple, published alongside the main text. He seeks to translate difficult Aramaic terms and help the reader understand the basic flow of the discussion. The bolded texts are quotes from the main text, followed by Rashi's words.
The Tosafot commentary is a compilation of analyses, explanations, and practical notes written by the Tosafists, who lived in the generations after Rashi. This commentary deals with difficult sections and contradictions within the main text and within the earlier Rashi commentary. It too became a staple of Talmud study.
The standard Vilna edition of the Talmud is divided into tractates, which are then divided into double-sided pages. Each daf (double-sided page), side a or side b. The Hebrew notation uses hebrew letters as numerals, followed by either a single dot (.) for side a or a colon (:). for side b. Interestingly, the first page of each tractate isn't 1; instead, they begin at 2a!
Why did you make this? Can't the Talmud already be found online?
Although the text of the Talmud and its commentaries can be found online, the texts alone without the visual form do not completely capture the traditional learning experience of the Talmud. Pragmatically, many Talmud classes reference the lines of the standard text during their exposition ("look at the third wide line"). But the visual form also has great cultural significance. Stories abound of great Talmudists who memorized the visual form and were able to pass the "pin test"—without looking, they could tell you each word that a pin passed through when pushed into the book. In traditional circles, the standard visual form is considered part of the tradition, along with the content.
To us aspiring digital humanists, the visual form of the Talmud represents not only its past, but its future. Although it is a product of the technological requirements of its time—cramming as much text in a square inch of printed paper as possible—it also serves as a platform for innovation that, rather than supplant the rich history of the daf, can continue it into the digital age.
We started asking user experience questions: what features, given the digital affordances of the web, could we add to the daf to make it easier to read? We also started thinking about the multivocality of the Talmud, the generations of scholars that it portrays. How could we represent the plurality of voices that study the Talmud today, including groups that were never included before? To begin to experiment and answer these questions, we needed a platform that not only implemented the daf in a digital format, but did it in an extendable way. The only existing way to access the Talmud with its visual form online is through a PDF or through implementations in proprietary formats, neither of which allow us to build off of it and add our own features. (Take for example, the Mercava, which, while free, is proprietary, closed-source, and provides no API or other way to extend it.) Our version is a native web interpretation, paving the way for ourselves and other developers to extend the daf using standard tools and technologies.
How did you do it?
Text on the web is built on a box model, making it a challenge to represent the Talmud's varying text shapes. We compare the lengths of the three texts on a given page, and use those lengths to create boxes of texts and cutouts of those boxes, pushing the text to flow in the desired, dynamically-determined shape. The only inputs to our algorithm are the three texts, and we infer the rest. The project is written in JavaScript. The texts come from Sefaria's API.
How close is it to the actual, Vilna print?
Most pages are pretty close, but there is much room for improvement. An algorithim attempting to predict the choices of a human editor can only go so far, especially since the only data we have access to are the texts themselves. For example, in the standard printed edition, there are pages where a long Tosafot comment will spill over from one page to another, but we don't have the data for when that happens. In the future, we plan on using image recognition to collect more data than is currently publically available, so that we can account for these special cases. We'll be implementing more algorithm cases and checks to ensure the text and its layout is as accurate as possible.
Who is this for?
Because our framework is modular and extendible, developers can use our renderer and algorithm to develop any sort of daf project. In addition to providing rendering options, our programming interface allows the developer to search, select, and manipulate words and sentences on the daf, opening a world of possibilities for features and extensions.
We have beginners, scholars, and avid Talmud learners in mind. We envision, for example, robust translation tools to help beginners understand the text, textual comparison tools to allow scholars to easily compare similar passages, and notetaking and networking tools to allow regular readers to save, take notes on, and share passages of the Talmud.
Who made this?
This project was conceived and developed by Dan Jutan and Shaun Regenbaum, two undergraduates at Georgia Tech, as part of the Digital Integrative Liberal Arts Center. They're graduates of Yeshivat Har Etzion and Yeshivat Birkat Moshe, respectively. Janet Murray and Brad Rittenhouse are faculty advisors on the project.
What's next for the project?
We'll be continuing to work on our algorithm and our codebase, and we plan to release an open-source library later this year. We also want to continue to build off the framework, to experiment with new features and create innovative learning platforms.
If the project is renewed for another semester, we hope to explore the democratization of Talmud learning by integrating the daf with the multitude of digital Talmud content that is created every day. We'd also like to bring 20th and 21st century Talmud scholars writing in English to the daf, allowing the user to click on a difficult passage and see new interpretations.