摘要
涉案新闻话题摘要的效果依赖于抽取的句子的质量,然而现有方法存在搜索空间较大、提取到过多非涉案和重复内容的问题。因此,提出了基于压缩空间句子选择的涉案新闻话题摘要方法,通过句子重要性评估模块过滤掉无关信息来压缩搜索空间,计算编码后文档集合中句子的突出特征和重复特征,平衡两种特征后提取出最终得分高的句子组成摘要。在数据集上进行的各项实验结果表明,所提方法可以生成高质量的涉案新闻话题摘要,并且各项评价指标均有提升。
The effect of the summarization of the case-related news topics depends on the quality of the extracted sentences.However,the existing methods have the problems of large search space and extracting too many non-involved and repetitive content.Therefore,a method for summarization of case-related news topics based on sentence selection in compressed space is proposed.The sentence importance assessment module filters out irrelevant information to compress the search space,calculates the prominent features and repetitive features of sentences in the encoded document set,and balances the two features to extract sentences with higher final scores to compose summaries.Various experimental results on the dataset indicate that the proposed method can generate high-quality summaries of case-related news topics,and all evaluation indicators are improved.
作者
卢天旭
LU Tianxu(Kunming University of Science and Technology,Kunming Yunnan 650500,China)
出处
《通信技术》
2022年第9期1136-1145,共10页
Communications Technology
关键词
涉案新闻
话题摘要
压缩空间
句子选择
句子重要性评估
case-related news
topic summarization
compressed space
sentence selection
sentence importance assessment
