当前位置: X-MOL 学术Neural Process Lett. › 论文详情
Our official English website, www.x-mol.net, welcomes your feedback! (Note: you will need to create a separate account there.)
A Topic Inference Chinese News Headline Generation Method Integrating Copy Mechanism
Neural Processing Letters ( IF 3.1 ) Pub Date : 2022-08-03 , DOI: 10.1007/s11063-022-10942-2
Zhengpeng Li, Jiansheng Wu, Jiawei Miao, Xinmiao Yu, Shuaibo Li

To maximize the accuracy of the news headline generation model, increase the attention ratio of the model to significant information, and avoid duplication of generated headlines and problems unrelated to feature semantics, we proposed a topic inference Chinese news headline generation method integrating a copy mechanism (TI-C-NHG). First, we enrich the TI-C-NHG input module to mine potential new topics through topic reasoning, making topic understanding a broader source of sentence-level context. Second, we propose a copy mechanism that can copy words from a vocabulary and news texts with topic information, which helps to improve the accuracy and readability of headings. In addition, the training model constructed by a multi-layer Transformer-Decoder can greatly improve the parallel ability of the model and speed up the inference process of headline generation. We verified the validity of TI-C-NHG in the Chinese Short Text Summary Datasets and the LCSTS datasets.



中文翻译:

一种结合复制机制的中文新闻标题生成方法

为了最大限度地提高新闻标题生成模型的准确性,提高模型对重要信息的关注率,避免生成的标题重复和与特征语义无关的问题,我们提出了一种结合复制机制的主题推断中文新闻标题生成方法( TI-C-NHG)。首先,我们丰富了 TI-C-NHG 输入模块,通过主题推理挖掘潜在的新主题,使主题理解成为更广泛的句子级上下文来源。其次,我们提出了一种复制机制,可以从带有主题信息的词汇和新闻文本中复制单词,这有助于提高标题的准确性和可读性。此外,多层Transformer-Decoder构建的训练模型可以大大提高模型的并行能力,加快标题生成的推理过程。我们在中文短文本摘要数据集和 LCSTS 数据集中验证了 TI-C-NHG 的有效性。

更新日期:2022-08-04
down
wechat
bug