Computational Interpretation of Embedded Narrative Discourse

Professor Pablo Gervás

Universidad Complutense de Madrid

Abstract:
The discourse for narratives beyond the simplest ones often conveys at different points of its span conflicting views on the events it is describing. The characters in the narrative will tell stories to other characters, and these stories are not always true. Authors will often exploit this to mislead someone--sometimes particular characters, sometimes the reader. In doing this, the author is hoping to achieve not a discourse intended to reflect the truth of the events at a given point in time, but rather an evolving sequence of views about those events presented in the specific order in which the narrator wants the reader to experience them. The discourse for a complete narrative will usually include several conflicting views in terms of what events are true, and these views need to be managed. Sometimes the reader needs to keep more than one view in mind. For instance, in the tale of Snowhite, the hunter tells the queen that he has killed Snowhite, but he has really let her go. Readers easily keep track of the beliefs of the different characters, and then adjust well when the queen discovers the truth and sets out to kill her herself. Sometimes they only have one view but it can change radically at a particular point in the story. In the Starwars saga, Luke Skywalker has been told that Darth Vader killed his father, and both he and the viewers believe that is so until the big reveal in The Empire Strikes Back. A process of interpretation of narrative would need to model how these various views are constructed, stored, and validated or falsified as the reader progresses through the reading of the discourse. Computational procedures for interpreting a story need to account for these embedded stories in terms of how to represent them and how to process them. The talk will present ongoing work towards the construction of a computational model that can identify embedded discourses, represent the way they are used to convey recursively embedded stories and interpret the resulting structures into a sequence of evolving interpretations of what is happening in the story.

Profile:
Pablo Gervás holds a PhD in Computing from Imperial College, University of London (1995), and he is currently a full professor of computational creativity and natural language processing (Catedrático de Universidad) at Universidad Complutense de Madrid. He is the director of the NIL research group (nil.fdi.ucm.es) and for many years he was the director of the Instituto de Tecnología del Conocimiento (www.ucm.es/itc). He has been the national coordinator for Spain of the FP7 EU projects PROSECCO, WHIM, and ConCreTe in the area of Computational Creativity. He has been the coordinator for two national research projects (GALANTE and MILES) involving several institutions and the principal investigator for two more (IDiLyCo and CANTOR). His main research interest currently lies in the study of the role that computers can play in helping people interested in literary creativity. Prof Gervás has expertise in automatic generation of (fictional) stories and poetry, and has a background in natural language generation, Computational Creativity, and narratology. He is the author of the PropperWryter software which was used in the process of creating Beyond the Fence -- the first computer-generated musical, staged at the London West End in 2016.

Week 23 2022/2023

Thursday 11th May 2023
1:00-2:00pm

Microsoft Teams - request a link via email

For queries or meeting link, contact Dr. Ignatius Ezeani (i.ezeani@lancaster.ac.uk)