ABSTRACT
Attempts to produce computer story understanding systems have generated a number of advances in the areas of knowledge representation and organization. However, many basic questions still remain largely unaddressed. In particular, the idea of what actually constitutes a story has never been clearly delineated.
A theory of story points has been developed that attempts to characterize those texts that describe situations that constitute stories. Unlike previous attempts at such a characterization, points are based not on the structure or form of a text, but on its content. Points describe those situations that generate reader interest and therefore give a text some poignancy.
The theory of stories proposed here is intimately connected with basic issues of language understanding, language generation, cognition and memory. In addition to characterizing stories, knowledge about poignancy is as necessary for the construction of intelligent story understanding programs as are theories of inference and knowledge representation. A story understanding system (PAM—Plan Applier Mechanism) has been given some of this knowledge about points, and as a result, its language processing capabilities have been extended to facilitate summarization and intelligent forgetting.
