Sequentially Controlled Text Generation

Abstract

While GPT2 generates sentences that are remarkably human-like, longer documents can ramble and are structurally different from human-written articles. 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, based on extensions of existing classifier-based approaches. We test different degrees of structural awareness and show that, in general, more structural awareness results in higher control-accuracy, grammaticality, global coherency and topicality, approaching human-level writing performance.

Tasks

Reproductions