Sequentially Controlled Text Generation

TL;DR

This paper introduces a new task for controlling the structure of long text generation, proposing a pipeline that improves coherence, grammaticality, and topicality by increasing structural awareness.

Contribution

It presents the concept of sequentially controlled text generation, a new dataset NewsDiscourse, and a generation-editing pipeline that enhances structured long-form text output.

Findings

Higher structural awareness improves control accuracy.

Enhanced structural awareness leads to more coherent and grammatical text.

Approaches approach human-level writing performance.

Abstract

While GPT-2 generates sentences that are remarkably human-like, longer documents can ramble and do not follow human-like writing structure. We study the problem of imposing structure on long-range text. We propose a novel controlled text generation task, sequentially controlled text generation, and identify a dataset, NewsDiscourse as a starting point for this task. We develop a sequential controlled text generation pipeline with generation and editing. We test different degrees of structural awareness and show that, in general, more structural awareness results in higher control-accuracy, grammaticality, coherency and topicality, approaching human-level writing performance.