Evolving stories: Tree adjoining grammar guided genetic programming for complex plot generation

Abstract

In this paper, we develop a tree adjoining grammar (TAG) to capture semantics of a story with long-distance causal dependency, and present a computational framework for story plot generation. Under this framework, TAG is derived and a story plot is represented by a derivation tree of TAG. The generated plots are then evolved using grammar guided genetic programming (GGGP) to generate creative, interesting and complex story plots. To evaluate these newly generated plots, a human-in-the-loop approach is used. An experimental study was carried out, in which this framework was used to produce creative, interesting and complex plots from a predesigned fabula based on a story known as "The magpie and the water bottle". The experimental study demonstrated that TAG and GGGP can potentially contribute significantly to complex automatic story plot generation. © 2010 Springer-Verlag.