Neural Sentence Ordering

TL;DR

This paper studies sentence ordering as an independent task, introducing a new dataset and pairwise learning approach, demonstrating its effectiveness through extensive experiments in natural language generation.

Contribution

It presents a large academic text corpus and a data-driven pairwise sentence ordering method, focusing solely on the ordering task rather than downstream applications.

Findings

Effective pairwise ordering model validated by experiments

Large dataset of academic texts for sentence ordering

Source code and dataset will be publicly available

Abstract

Sentence ordering is a general and critical task for natural language generation applications. Previous works have focused on improving its performance in an external, downstream task, such as multi-document summarization. Given its importance, we propose to study it as an isolated task. We collect a large corpus of academic texts, and derive a data driven approach to learn pairwise ordering of sentences, and validate the efficacy with extensive experiments. Source codes and dataset of this paper will be made publicly available.