Improving Pacing in Long-Form Story Planning
Yichen Wang, Kevin Yang, Xiaoming Liu, Dan Klein · 2023
Existing LLM-based systems for writing longform stories or story outlines frequently suffer from unnatural pacing, whether glossing over important events or over-elaborating on insignificant details, resulting in a jarring experience for the reader.We propose a CONCrete Outline ConTrol (CONCOCT) system to improve pacing when automatically generating story outlines.We first train a concreteness evaluator to judge which of two events is more concrete (low-level-detailed).This evaluator can then be used to control pacing in hierarchical outline generation; in this work, we explore a vaguest-first expansion procedure that aims for uniform pacing.We further use the evaluator to filter new outline items based on predicted concreteness.Compared to a baseline hierarchical outline generator, humans judge CONCOCT's pacing to be more consistent over 57% of the time across multiple outline lengths; the gains also translate to downstream stories.All code, data, and models are open-sourced.1 Outline GenerationCONCOCT uses our concreteness evaluator M to improve outline pacing in two ways: vaguest-first expansion order and concrete candidate generation.High-Level Outliner Structure.We view a hierarchical outline as a tree, rooted at the overall 10789 Chapter-Level Paragraph-Level Split Size Summary Len Raw Len Raw / Sum Size Summary Len Raw Len Raw / Sum Train 23,564 133.7 5450.7 40.77 162,122 58.6 71.6 1.22