This talk will use science fiction, fantasy, mystery, and the Gothic to explore the advantages of an approach that asks data science to contribute to the humanities by adding perspectival flexibility, rather than sheer scale. Underwood trained predictive models of these genres using ground truth drawn from various sources and periods (19c reviewers, early 20c bibliographies, contemporary librarians), in order to explore how implicit assumptions about genre consolidate or change across time. Contrasting different models also allows us to take a parallax view of individual books, or even paragraphs in books: which passages in 1960s science fiction, for instance, would have been hardest for a pre-war reader to recognize as SF?
Ted Underwood teaches in the School of Information Sciences and the English Department at the University of Illinois, Urbana-Champaign. He was trained as a Romanticist and now applies machine learning to large digital collections. His most recent book, Distant Horizons: Digital Evidence and Literary Change (Univ of Chicago, Spring 2019) addresses new perspectives opened up by large digital libraries.
Find information about the related seminar here.
This event is part of a 3-part Fall 2018 DH lecture series. Find more information about the full series here.