A handle on the stuff that never gets read

“I am passionate about literature and have an interest in mathematics and statistics. The nice thing is that the days are long gone when you had to pursue these different interests in separate worlds. Thanks to the digital humanities and, more recently, computational literary studies, we now have a wonderful community in which the study of literature can be combined with computer-assisted and empirical work. In this field, I want to play a part in opening up and bringing together new and classical philological perspectives on literature and literary history.”

As early as his school days, Julian Schröter discovered his affinity for both literature and technology. German and math thus became his main subjects. At the University of Würzburg, he then studied philosophy and German literary history. He later also earned a doctorate in German literature and literary theory at the same university before commencing work on his professorial qualification on the subject of “Esthetic and social functions of stories and novellas in the 19th century. Literary history as mixed methods research”. A Walter Benjamin Fellowship granted by the German Research Foundation (DFG) took him to the Universities of Antwerp and Illinois Urbana-Champaign. A position as Deputy Professor for Digital Humanities at the University of Trier preceded his move to LMU in 2023.

Digital methods supported his research in two ways. First, it was these methods that made large volumes of text available in the first place. “In various projects, I digitalize literature and tap into metadata about the context surrounding their media history or the history of the material covered. In the past, such classical philological work was very painstaking.” Now, quantitative, computer-assisted literary studies opens up a very broad view of literature that can also factor in the social, economic and technological circumstances of a given era – such as changes in printing technologies.

“Coping with huge volumes of text”

Another strength of the new technologies, the professor adds, is the ability to read large bodies of text, trawl through them for ever more complex esthetic phenomena and apply statistical methods to provide an overview of distributions: “Since the end of the 19th century at the latest, there has been a very large market that has still never been properly surveyed and that human readers can simply never master – the result being that people repeatedly read the same standard works.” Modern methods, he says, deliver a rough outline of all the stuff that never gets read and make it easier to compare texts. “That gives us better ways to assess the high-level literary canon and its relationship to mass culture.”

The ‘perspective modeling’ technique trains artificial intelligence (AI) to assign texts to 19th-century literary categories. The models that emerge from this exercise can be shifted along the timeline in simulation runs – from the period around 1800 to the time around 1900, for example. “Literary history can be sketched like a long-term study with major horizon lines that philologists have never seen before.”

In this way, it is possible to shed light on shifts in semantics or in the way generic terms are used – and, for example, to discuss whether defined classification systems influenced the way texts were written. “Beyond that, new questions arise: Aside from its length, is a novel really different to a story or novella, or are many of the distinctions from which we derive categories ultimately nothing more than ideological constructs?”

“Simulating the forgotten canon”

In one current project, Schröter wants to use AI to operationalize the recognition of how suspense is perceived in texts. “Suspense is a psychological phenomenon that translates only indirectly into written texts,” he explains. “But the latest language models can identify it with the aid of textual indicators – dangerous situations, say, and the protagonists’ fear.”

This approach allows the phenomenon of suspense to be analyzed form a number of angles, he says; and it works for the recognition of such varying aspects as types of landscape, conflict situations and the scene of events. Quantitative analysis and the precise reading of literature can be combined – in teaching contexts, too – to wonderful effect: Looking carefully at the way computer-assisted methods work and the results they produce can thus give us a better understanding of established philological and narrative analysis concepts such as suspense, plot and the character constellation. Close and distant reading can thus effectively cross-fertilize each other.

“Within the discipline of literary studies, opinions vary on when the cliffhanger came into being,” Schröter says. “Some say it began with Arabian Nights, others point to 20th-century TV series. Quantitative methods are a very good way to answer these questions across the spectrum of the popular mass market. In the course of the 19th century, there was a marked increase in serial publishing – the practice of breaking novels and narrative prose down into lots of individual episodes. This phenomenon created ideal conditions for the evolution of the cliffhanger. That said, it is possible to show that it took a fairly long time before use of the cliffhanger became commonplace in the course of the 19th century. Computer-assisted analytical techniques can also be tremendously useful regarding the major issue of how continuous or disruptive cultural change has taken place.

Additionally, mathematical models help us estimate the rough magnitude of underpinning bodies of text that have already been lost from a given epoch. “In the future, large language models could hypothetically reconstruct aspects of historic reality – effectively as simulations – in such a way that we can better envisage the rules by which literary and other works were passed on or, indeed, not passed on but abandoned and forgotten – or even destroyed.”

Alongside the historical analysis of texts using AI, Schröter is also interested in literature that is produced by AI, “especially with a view to interaction between humans and language models”. In the last summer semester, he ran a seminar on ChatGPT in the context of literary studies. Right now, he and a number of colleagues are currently focusing in particular on the digital humanities, addressing the issue of whether and how language models create textual meaning or perhaps only simulate it.

Coding as in musicology

This research has opened up all kinds of overlaps with other disciplines. Musicology, for instance, codes its works using the same XML format as our texts,” the literary scholar notes. “Thanks to common data standards, we can, for example, analyze cultural change on a broader basis and in a less isolated manner.” Digital editions, which are especially important for the humanities, are making close collaboration between philology, musicology and art history ever more important.

The CIS – a “legendary center for computer linguistics” – and the Humanities IT group create optimal conditions for Schröter’s research at LMU. The professor nevertheless sees distinct profiles emerging within the digital humanities: “Digital archeology, history and musicology stand out more sharply as independent disciplines with their own conferences, journals and communities. It is great to see such huge interest in interdisciplinary and transdisciplinary collaboration between the various aspects of the humanities and cultural science at LMU.” Numerous cooperative ventures are already underway. Digital methods are thus one of many ways to bring different disciplines together.”

Schröter admits that certain reservations still exist: “That,” he says, “is no surprise, because there has always been a suspicion that empirical methods are far removed from reading and from ‘proper’ literary studies, whatever that means.” Yet he believes it is not about replacing certain methods with other methods, stressing that people will always work with the methods and techniques that suit them: “I believe that the future of literary studies will consist of people with different skills working together. I myself thoroughly enjoy familiarizing myself with new technologies and understanding their mathematical foundations. So, researchers with an affinity for new technologies and empirical methods can forge valuable links between the disciplines.”