TNA Fellow Roxana Patras: Criss-crossing Novel Illustrations and DALL-E-generated Images: toward a data-rich ekphrasis
The project aims to bridge latest technology (DALL-E image generator) with historical data: a. descriptive sequences of novels (faces and places); b. illustrations of novels. Committed to supplement the existing textual resources, the project also involves a reconsideration of the original artefact that is timely and necessary in CLS: the textualism’s legacy made text-processing prevail over image processing. The interest in images is spurred by the ascertainment that lesser resourced languages are difficult to integrate in multilingual investigations unless they address universalia such as visual devices. By “data–rich ekphrasis”, the output of the illustrating automaton when fed in with sequences of words and images linked in or belonging to the same plot is understood. Generally speaking, it is a machine replication of the reader’s response to a set of textual and visual characteristics aided by the reader’s memory and engagement; it is the machine’s capacity to “paint with words”, to draw images from the literary text and from illustrations of the literary text, by using pre–trained pictorial material. The exploration of ekphrasis develops from the hypothesis that the illustrations of the consumer novels are, like the texts themselves, aggregations of (visual) cliches. Thus, the confrontation of human illustrations and automatic illustrations (produced with DALL-E fed in with description chunks) might shed light on the functionality of the cliché and of the ready-made sequences.
Roxana Patras works at the Institute for Interdisciplinary Research, “Alexandru Ioan Cuza” University of Iasi. She does research in History of Literature, Literary Theory, Rhetoric, and Computer Literary Studies. Her current research interest is 19th-century Romanian Popular Fiction and emerging novel genres.