It must be kept in mind that this is the role that the ISAAC model fills in my research. When reviewing my theory, one must be careful to separate those aspects which rightfully belong to the theory from those aspects which belong to this particular computer instantiation of the theory. For example, in a later paragraph, the technique by which the dynamic changing of a concept's function is accomplished in the ISAAC system is explained, namely, via spreading activation. The theory itself does not make a commitment to spreading activation, it merely states that this ability to reconsider the function of a concept is necessary. The implementation is where the tie to the spreading activation technique is realized. Again, it is important to keep the separation of the reading theory and its implementation in the ISAAC model in mind as we progress. Otherwise, the theory will be misrepresented by adding to it features which rightfully belong in the computer model alone. Likewise, some of ISAAC's current limitations are limitations of the instantiated model rather than limitations of the theory itself.
ISAAC reads two types of stories. First, it can read multipage science fiction stories of which there are five currently handled by the system. Second, one to three paragraph synopses of Star Trek stories have been used; the system can read nine of these. If new stories were to be added to the range of ISAAC, concepts would need to be added to the background knowledge. As I said in Chapter 6, understanding can only be successfully if there is some link to pre-existing knowledge. The stories which are read are previously published stories and synopses which have appeared in print. While I have chosen the stories, my choice was based on example stories which would demonstrate interesting aspects of creative understanding, rather than on stories which were just ``simple'' enough to be read. The short stories (which appear along with the synopses in Appendix A) are:
ISAAC is implemented in Common LISP. As the core of the knowledge representation system, the KR frame package ([#!frame:kr!#]) is used; the package which KR is part of, Garnet, also provide graphical interface components to the system, although I have not stressed this aspect of the model. With respect to the supertasks, each of the six supertasks is implemented to the level necessary to support the creative understanding process. Some of the supertasks, though, are implemented by building on other research projects. For example, ISAAC currently uses an extension of the COMPERE parsing system ([#!sentence:kavi1!#]) as an implementation for its sentence processing supertask. Similarly, the ISAAC system makes use of the MOORE memory system ([#!memory:ram-francis-1996!#]) to provide the functionality of the memory supertask.