While Transformers have had significant success in paragraph generation, they
treat sentences as linear sequences of tokens and often neglect their
hierarchical information. Prior work has shown that decomposing the levels of
granularity~(e.g., word, phrase, or sentence) for input tokens has produced
substantial improvements, suggesting the possibility of enhancing Transformers
via more fine-grained modeling of granularity. In this work, we propose a
continuous decomposition of granularity for neural paraphrase generation
(C-DNPG). In order to efficiently incorporate granularity into sentence
encoding, C-DNPG introduces a granularity-aware attention (GA-Attention)
mechanism which extends the multi-head self-attention with: 1) a granularity
head that automatically infers the hierarchical structure of a sentence by
neurally estimating the granularity level of each input token; and 2) two novel
attention masks, namely, granularity resonance and granularity scope, to
efficiently encode granularity into attention. Experiments on two benchmarks,
including Quora question pairs and Twitter URLs have shown that C-DNPG
outperforms baseline models by a remarkable margin and achieves
state-of-the-art results in terms of many metrics. Qualitative analysis reveals
that C-DNPG indeed captures fine-grained levels of granularity with
effectiveness.

本文提出了一种连续分解的细粒度建模方法，名为 C-DNPG，并在句子编码过程中引入了 GA-Attention 机制，用于提高神经释义生成的效果。该方法在 Quora 问题和 Twitter 链接两个基准测试中都取得了显著的优势和最佳效果，并且在 qualitive 分析中表现良好。