Automated Story Generation as Question-Answering

TL;DR

This paper introduces a novel story generation method that frames the task as a question-answering process, starting from the story's end and iteratively generating preceding events to improve coherence and goal-directedness.

Contribution

It presents a new approach to automated story generation by modeling it as a generative question-answering task, addressing coherence and goal-orientation issues of previous neural methods.

Findings

Effective generation of coherent stories from the ending

Improved goal-directed story coherence

Demonstrates the viability of question-answering for story generation

Abstract

Neural language model-based approaches to automated story generation suffer from two important limitations. First, language model-based story generators generally do not work toward a given goal or ending. Second, they often lose coherence as the story gets longer. We propose a novel approach to automated story generation that treats the problem as one of generative question-answering. Our proposed story generation system starts with sentences encapsulating the final event of the story. The system then iteratively (1) analyzes the text describing the most recent event, (2) generates a question about "why" a character is doing the thing they are doing in the event, and then (3) attempts to generate another, preceding event that answers this question.