基于子句单元的异构图网络抽取式文本摘要

Extractive Text Summarization with Heterogeneous Graph Network Based on Sub-sentence Unit

的目标是将长文本进行压缩、归纳和总结,从而形成具有概括性含义的短文本,其能帮助人们快速获取文档的主要信息。当前大多数的抽取式文本摘要的研究都是以整句作为抽取单元,而整句作为抽取单元会引入冗余信息,因此该文考虑使用粒度更细的抽取单元。已有研究表明,细粒度的子句单元比整句单元在抽取式摘要上更具有优势。结合当下热门的图神经网络,该文提出了一种基于子句单元异构图网络的抽取式摘要模型,有效融合了词、实体和子句单元等不同层次的语言信息,能够实现更细粒度的抽取式摘要。在大规模基准语料库(CNN/DM和NYT)上的实验结果表明,该模型产生了突破性的性能并优于以前的抽取式摘要模型。

The goal of text summarization is to summarize long text into a short text with main information.To avoid the redundant information brought by the sentence extraction,we propose an extractive summarization model based on a heterogeneous graph network of sub-sentence units,which effectively integrates different levels of language information such as words,entities,and sub-sentential units.Experiments on two large scale benchmark corpora(CNN/DM and NYT)demonstrate that our model yields ground-breaking performance and outperforms previous extractive summarizers.