On the Evolving Transformation System Model Representation of Fairy Tales

This thesis presents an initial application of the Evolving Transformation System (ETS) formalism to the structural representation of Russian fairy tales. This work was motivated by the needs of information retrieval (IR) in general, but specifically concentrates on proposing a new type of document representation. The new representation is based on the temporal or constructive process of events that occur in the mind of the “generic” fairy tale listener/reader. Representing documents in this way yields several advantages to using ETS over traditional IR techniques, such as being able to retrieve partial information from a document (i.e. a single sentence or paragraph), fast searching by traversing the multilevel hierarchial class structure generated by the ETS intelligent process, and locating similar documents based on structural similarity rather than key words. Moreover, the ETS formalism suggests an explanation for the mechanism behind plot anticipation by an experienced reader as well as various levels of fairy tale conceptualization. Using the ETS formalism and in-depth analysis of many fairy tales, we propose approximately 40 ETS primitives that make up the underlying structure of events for i any Russian fairy tale. Using these primitive events, we develop the structural representation of several story segments belonging to the same class, which we then use to identify ETS transformations. Using these transformations, we develop the class representation and demonstrate the ETS multi-level representation for this class of fairy tale segments.