Knowledge packets

    The three primary supertasks act together to produce the explicit output structures which act as the ``result'' of a reading   experience--control,   story structure comprehension, and   scenario comprehension. To a lesser extent,   sentence processing also produces an explicit structure, although not at the level of the primary structures. Because of this structure construction, these four supertasks need additional information in order to function correctly. Specifically, these supertasks need to know about the kinds of structures that they may build. For example, a first-person narrative should be represented in a structure which is different in some ways from a third-person narrative.   These collections of knowledge governing the types of structures which can be built are referred to in my research as packets. Packets are just a specific set of concepts in the representational background of the reader; they contain the information which enables the reasoner to handle different reading situations. There is a direct relationship between the packets and the ultimate output representations being built by the reader. For example, consider a first-person narrative text. This should be represented by an output structure which captures the fact that it is indeed a story being told in the first-person. However, in order to take advantage of the fact that a first-person narrative is being dealt with, a reader must be able to recognize a text as falling into this category. To do this, the reader needs the first-person narrative packet, which acts a generic model of what a first-person narrative should contain. This abstract information is used to guide the reading process while building the story structure representation which captures the specific details of this particular first-person narrative. Each of the   primary supertasks and   sentence processing has a set of packets associated with it. The packets organize the detailed representations (using Chapter 4 formalism) into meaningful functional structures for the different supertasks to make use of while building their output representations.