多文档文摘提取方法的研究  

Investigation of method for extracting multi-document abstracts

作　　者：张其文[1] 李明[1] 

机构地区：[1]兰州理工大学计算机与通信学院,甘肃兰州730050

出　　处：《兰州理工大学学报》2007年第1期96-99,共4页Journal of Lanzhou University of Technology

基　　金：甘肃省自然科学基金(ZS022-A25-035);甘肃省教育厅科研基金(0416B-04)

摘　　要：在深入分析当前流行的自动文摘技术和方法的基础上,将语义方法融入统计算法,提出了一种基于统计的多文档文摘提取方法,描述了它的实现过程.该方法利用文档内和文档之间段落的语义相关性,实现了多文档文摘的自动生成.首先对文档进行分段实现信息分割,再利用聚类技术对文本段进行聚类实现信息合并,最后从类中抽取代表段产生摘要文本.实验结果表明,该方法是一个有效、实用的方法.Base on the analysis of currently prevalent techniques and methods of automatic abstracting thoroughly, a method for extracting multi-document abstracts was presented based on statistics, and its implementation process was described, also. By using semantic correlation among paragraphs in an article and that among articles, automatic generation of multi-document summary was implemented. First of all, texts were segmented into semantic paragraphs to implement information segmentation, and then, by using clustering algorithm, semantic paragraphs were merged into topic classes. Finally, representative paragraphs were extracted from classes to generate text summary and implement information condensation. Experimental results indicated that the method was effective and applicable in practice.

关 键 词：自动文摘 文本分段 语义段 向量空间模型 文本模型 

分 类 号：TP391[自动化与计算机技术—计算机应用技术] TP18[自动化与计算机技术—计算机科学与技术]