Ted Underwood: Predicting the Past

"Predicting the Past: How Digital Libraries Speak to Literary Questions about Genre, Gender, and Prestige"

Many scholars sense that digital libraries represent a research opportunity for the humanities. But the opportunity has been hard to grasp: the patterns that are easiest to mine digitally (word frequencies, network graphs, topic models) don't always clearly address existing disciplinary questions.

This talk reflects on a strategy for building bridges between humanities disciplines and information science that has worked well in several recent projects. The strategy poses only modest technical challenges: it starts with existing questions about the literary past, and addresses those questions with supervised learning algorithms that have long been used for information retrieval. But simple methods can be enlightening, if we understand the complementary strengths of humanistic interpretation and predictive modeling. I'll explain how these methods can illuminate the differentiation of genres, the slow transformation of literary prestige, and the waxing and waning of gendered assumptions about character.

Speaker bio: Ted Underwood studies literary-historical patterns that become visible across long timelines, especially in the last three centuries of English-language poetry and fiction. He is the author of two books, including Why Literary Periods Mattered (Stanford, 2013), and of articles in journals ranging from PMLA to Big Data and Society. A professor of English at Illinois, he also participates in interdisciplinary research groups centered at the University of Chicago and McGill University, and collaborates with HathiTrust Research Center.