To illustrate this, consider a ``normal'' reading experience for the ISAAC system. Empirical considerations led me to determine that a third-person narrative should act as the default story structure packet, if no other information is known prior to the start of a reading experience. This is not an unreasonable assumption as this genre of text is highly pervasive across stories with which a system may deal. The specific third-person narrative packet is further developed with some additional information. This includes:
Suppose that ISAAC begins reading a story which is actually a first-person narrative. In this case, at some point in the reading, the fact that a first-person pronoun is used outside the scope of a direct quotation will indicate to the system that a change in story structure packet is necessary. This pronoun usage is a trigger for the context switch. When this occurs, the first-person narrative packet is retrieved from memory. Information which has been discovered from the point of view of the incorrect third-person view will be preserved; e.g., characters which have been discovered will be transferred. The default first-person narrative packet is closely related to the third-person narrative form with one important difference--the protagonist is tagged as being the character identified as the narrator. This fact greatly aids the identification of the protagonist.
In either of these cases, the most notable exception is the lack of any information regarding the plot which the story contains. While this is an aspect of the story structure, the actual details of this is provided by the scenario comprehension supertask. Thus, for the narrative forms being considered in this research, the idea of plot is identical to the set of events which occurs in the story.
As the ISAAC system builds a history of reading stories, particular story structure packets can be generated. In this research, however, the only example of this is in the case of the Star Trek synopses which are used by the model. ISAAC creates a story structure packet which describes the generic Star Trek story. The most important feature of this is in terms of the protagonist--the main sympathy character of the Star Trek stories is almost always Captain Kirk. Once this fact is ``realized'' by the system, future synopses are read more rapidly as protagonist identification is no longer a required task. This is an extremely simple example; indeed, the technique which is used to build the generic packet is a simple abstraction process which finds commonalities between specific story reading episodes and stores a generic, abstract representation. This idea of building specific story structure packets in order to handle similar stories is a powerful one--for example, a reader may carry around story structure packets for Zane Grey westerns, Heinlein juvenile novels, Star Trek stories, and so on, thereby allowing highly specialized reading behavior to occur--which, unfortunately, falls outside the scope of the current implementation. It is a prediction which the theory makes, but I did not explore it fully. The idea, however, is one direction for future research to take.
