Narrative Text Summarization

Abstract

In order to summarize a story it is necessary to access a high level analysis that highlights the story's central concepts. A technique of memory representation based on affect units appears to provide the necessary foundation for such an analysis. Affect units are conceptual structures that overlap with each other when a narrative is cohesive. When overlapping intersections are interpreted as arcs in a graph of affect units, the resulting graph encodes the plot of the story. Structural features of the graph then reveal which concepts are central to the story. Affect unit analysis is currently being investigated as a processing strategy for narrative summarization. When a reader summarizes a story, vast amounts of information in memory are selectively ignored in order to produce a distilled version of the original narrative. This process of simplification relies on a global structuring of memory that allows search procedures to concentrate on central elements of the story while ignoring peripheral details. It is apparent that some hierarchical structure is holding memory together, but the precise formulation of this structure is much more elusive. How is the hierarchical ordering of a memory representation constructed at the time of understanding? Exactly what elements of the memory representation are critical in building this structure? What search processes examine memory during summarization? How are summaries produced after memory has been accessed? In this paper we will outline a strategy for narrative summarization that addresses each of these issues. This proposed representation for high level narrative analysis relies on affect units. An affect unit is an abstract structure composed of three affect states and four affect links.