利用维基百科实体增强基于图的多文档摘要.pdfVIP

中国科技论文在线 利用维基百科实体增强基于图的多文档 摘要# 陈维政，严睿，闫宏飞，李晓明** 5 （北京大学网络与信息系统研究所，北京 100871） 摘要：针对基于图的多文档摘要，该文提出了一种在图排序中结合维基百科实体信息增强摘 要质量的方法。首先抽取文档集合中高频实体的维基词条内容作为该文档集合的背景知识， 然后采用 PageRank 算法对文档集合中的句子进行排序，之后采用改进的 DivRank 算法对文 档集合和背景知识中的句子一起排序，最后根据两次排序结果的线性组合确定文档句子的最 10 终排序以进行摘要句的选取。在 DUC2005 数据集上的评测结果表明本文的方法可以有效利 用维基百科知识增强摘要的质量。 关键词：自然语言处理；多文档摘要；维基实体；基于图 中图分类号：TP31 15 Using Wikipedia Entities to Enhance Graph-based Multi-document Summarization Chen Weizheng, Yan Rui, Yan Hongfei, Li Xiaoming (Computer Networks and Distributed Systems Laboratory, Peking University, Beijing 100871) Abstract: This paper presented a novel method to enhance graph-based multi-document 20 summarization by incorporating Wikipedia entities information. The Wikipedia pages’ content of high-frequency entities were extracted and arranged as the document collections’ background knowledge. After that the PageRank algorithm was used to sort these sentences in the document collections. Then an improved DivRank algorithm was used to sort the sentences both in the document collections and the background knowledge. Finally the summary sentences were chosen 25 based on their final rank which is a liner combination of these two ranking results. Results of experiments on the data of document understanding conference (DUC) 2005 show that the method proposed in this paper can effectively make use of the Wikipedia knowledge to improve