The article presents a novel approach to sentence ordering called the Non-Autoregressive Ordering Network (NAON). Existing sentence ordering approaches generally use an autoregressive manner that cannot fully explore the semantic dependency between sentences for ordering. The NAON, however, is able to explore bilateral dependencies between sentences and predicts the sentence for each position in parallel. This approach is suitable for sentence ordering due to two unique characteristics of the task: deterministic length of each generation target, and exclusive match between sentences and positions. The authors demonstrate that the NAON outperforms all the autoregressive approaches and yields competitive performance compared with the state-of-the-arts.
Publication date: 20 Oct 2023
Project Page: https://github.com/steven640pixel/nonautoregressive-sentence-ordering
Paper: https://arxiv.org/pdf/2310.12640
