Internet网络新闻文本自动摘要的研究_官礼和.pdfVIP

第28 卷 第14 期 计算机工程与设计 2007 年7 月 Vol . 2 8 N o . 14 Computer Engineering and Design July 2007 Internet 网络新闻文本自动摘要的研究 官礼和 ( 重庆交通大学信息与计算科学研究所，重庆400074) 摘 要：给出了Internet 网络新闻中文文本自动摘要的基本思路和基本步骤，讨论了断句、分词算法。针对自动摘要中新闻文 本的4 种形式特征，提出了一套新的自动摘要方案：首先综合新闻文本的4 种形式特征对词汇和句子赋予不同的权值，然后根 据权值大小按给定的比例挑选句子, 并进行平滑处理，生成文字流畅且具备一定质量的摘要。最后实验分析表明效果较好。 关键词：自然语言处理; 自动摘要; 分词; 断句; 权重; 平滑处理 中图法分类号：TP319 文献标识码：A 文章编号：1000-7024 (2007) 14-3518-03 Research on automatic abstract of news text of Internet network GUAN Li-he (Institute of Information and Calculation Science, Chongqing Jiaotong University, Chongqing 400074, China) Abstract ：The basic thinking and step of the automatic abstract of the news text is provided, and the algorithm of words frequency and breaking sentences is discussed. To contrapose 4 kinds of form characteristics of the news text in the automatic abstract, a set of new automatic abstract schemes are put forward: First, using these form characteristics, different weights are given to the vocabulary and sentence. And then, some sentences are selected according to the proportion definitely and the size of weights. Moreover, dealing with them smoothly. And last, generating automatic abstract, which is of some quality and fluent. Key words ：natural language processing; automatic abstract; words frequency; breaking sentences; weights; dealing with levelly and smoothly 吴岩、王开铸等早在1998 年就对英文自动摘要进行了研 0 引 言 究，并开发了HIT-971 型英文自动文摘系统，但摘要质量仍需