Abstract
Narrative understanding involves determining what matters in a story, maintaining coherence over extended stretches of text or video, and tracking how expectations evolve over time. Explaining this process in humans, and building natural language processing systems that can model it, is a central challenge for cognitive science and AI. In this talk, Frank Keller presents computational approaches to narrative salience, suspense, and plot structure in text and film. He shows how these models support long-range inference in complex narratives while also providing formal accounts of human narrative processing. More broadly, this work aims to develop narrative models that are both computationally effective and cognitively meaningful.
Speaker Biography
Frank Keller is professor of computational cognitive science in the School of Informatics at the University of Edinburgh. His research bridges natural language processing, computer vision, and cognitive science, with a focus on multimodal AI, long-form narrative understanding, and computational models of human language processing. Keller has held visiting appointments at the Massachusetts Institute of Technology, the University of Washington, and New York University, and has served in major leadership roles at Edinburgh, including as deputy head of school and training director for doctoral programs in NLP and responsible AI. He is a European Laboratory for Learning and Intelligent Systems Fellow and recipient of an European Research Council Starting Grant and a Leverhulme International Fellowship.