Abstract by Michael DeBuse
Plot Extraction and the Assessment of Event Importance in Stories
Plot extraction is presented as a literary analysis tool different from event extraction, requiring causal relationships and assessment of the importance of events in a story text. I show a naive method of event segmentation using the mean shift algorithm as a method of determining areas of potential plot importance, discuss its shortcomings, and propose a better method. This method uses causation, where the story text is ingested in streaming fashion--starting from the beginning without needing to know the story in completion--and events are appended to a narrative-ordered graph structure where unnecessary nodes are trimmed if no causal relationship is shown with future events.