Integrating Purpose and Revision into a Computational Model of Literary Generation

Abstract

Over the past few years, advances in the area of computational creativity have explored the combination of generative processes and evaluation models to obtain artifacts that are both original and valuable. This has taken place in fields as different as music, graphical art, or mathematics. Many of these efforts rely on identifying a sweet spot where an existing technology provides a constructive process that resembles closely some part of a creative process of humans. However, these approaches very rarely model the iterative nature of creative process as observed in humans, where a creator sets out with a purpose in mind, and creates drafts and revises them successively until the purpose is met. While it is a fact that in truly creative processes the purpose may also evolve during revision, the lack of direction has always been a damaging criticism to early approaches to computational creativity. In this chapter we will address a set of examples of approaches to the computational generation of literary texts based on particular techniques, propose a computational model that captures the purpose-driven revision of generated artifacts, and review a number of ongoing research efforts that implement parts of this model.