基于多特征和Ranking SVM的微博新闻自动摘要研究  

Micro-blog-oriented Chinese News Summarization Based on Multi-feature and Ranking SVM Algorithm

作　　者：李孟爽 昝红英[1] 贾会贞 

机构地区：[1]郑州大学信息工程学院,河南郑州450001

出　　处：《郑州大学学报（理学版）》2017年第2期43-47,共5页Journal of Zhengzhou University:Natural Science Edition

基　　金：国家自然科学基金项目(61402419);国家社会科学基金项目(14BYY096);国家高技术研究发展863计划项目(2012AA011101);国家重点基础研究发展计划973课题(2014CB340504)

摘　　要：提出了面向微博应用的新闻文本自动摘要研究方法.利用互信息对新闻文本中词语和句子之间的语义特征进行计算,根据其关联度对句子进行主题划分,赋予主题句较高的权重,同时从文本中抽取多种组合特征,利用Ranking SVM对句子进行排序,从而得到自动摘要.在NLP&CC2015面向微博中文新闻自动摘要评测数据集上进行对比实验,取得了良好效果,证明该方法的有效性.A Micro-Blog-oriented Chinese news automatic summarization algorithm was proposed.Mutual information was used to compute semantic feature among words and sentences in the Chinese news text.Then different topics were divided according to the correlation among the sentences,and a higher weight was gaven to topic sentences.Various combined features were extracted from the text,and sentences were ranked by Ranking SVM algorithm.The result of this algorithm demonstrated the effectiveness of the method.

关 键 词：互信息 语义特征 主题句 新闻文本自动摘要 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]