arXivDaily arXiv每日学术速递 周一至周五更新
arXiv 2102.00385cs.CL
收藏

文本摘要的语境化重写

Contextualized Rewriting for Text Summarization

Westlake University(西湖大学);Westlake Institute for Advanced Study(西湖高等研究院)

Guangsheng Bao, Yue Zhang

摘要:

摘要存在着不相关、冗余和不连贯等问题。已有的工作表明,摘要的抽象重写可以提高摘要的简洁性和可读性。这些重写系统将提取的摘要视为唯一的输入,这是相对集中的,但可能丢失重要的背景知识。在本文中,我们研究语境重写,它摄取整个原始文档。我们将上下文重写形式化为具有组对齐的seq2seq问题,引入组标签作为对对齐进行建模的解决方案,通过基于内容的寻址来识别提取的摘要。结果表明,我们的方法显着优于非上下文重写系统,而不需要强化学习,实现了多个提取摘要的ROUGE分数的大幅改善。

英文摘要:

Extractive summarization suffers from irrelevance, redundancy and incoherence. Existing work shows that abstractive rewriting for extractive summaries can improve the conciseness and readability. These rewriting systems consider extracted summaries as the only input, which is relatively focused but can lose important background knowledge. In this paper, we investigate contextualized rewriting, which ingests the entire original document. We formalize contextualized rewriting as a seq2seq problem with group alignments, introducing group tag as a solution to model the alignments, identifying extracted summaries through content-based addressing. Results show that our approach significantly outperforms non-contextualized rewriting systems without requiring reinforcement learning, achieving strong improvements on ROUGE scores upon multiple extractive summarizers.

机构、总结、中文摘要均由AI分析而来,可能存在错误,请以论文正文为准。

↑