基于HowNet概念获取中文自动文摘系统_王萌2004.pdfVIP

19 3 JOURNAL OF CHINESE INFORMATION PROCESSING Vol19 No3 : 1003- 0077 ( 2005) 03- 0087- 07 HowNet 1 1 2 1 , , , ( 1 , 430079; 2 , 119672) : 本文提出了一种中文自动文摘的方法不同于其它的基于词频统计的一般方法, 运用概念( 词义) 作 特征取代词语用概念统计代替传统的词形频率统计方法, 建立概念向量空间模型, 计算出句子重要度, 并对句子进行冗余度计算, 抽取文摘句对于文摘测试, 采用两种不同的方法进行测试: 一是用机器文摘和专 家文摘进行比较的内部测试; 二是对不同文摘方法进行分类, 通过对分类正确率的比较的外部评测方法 : 计算机应用; 中文信息处理; How Net ; 自动文摘; 概念向量空间模型 :TP391 : A Automatic Chinese Te t Summarization System Based on Conceptual Vector Space Model 1 1 2 1 WANG Meng ,HE T ng- t ng , JI Dong- hong , WANG X ao- rong ( 1. Central Ch na Normal Un vers ty Department of Computer Sc ence , Wuhan 430079, Ch na; 2. Inst tute for Infocomm Research Heng Mu Keng Terrace, 2 1, 119672, S ngapore. ) Abstract: The paper presents an approach for Ch nese text summar zat on . Unl ke normal stat st cal method, we use concept ( w ord sense) as feature, nstead of w ord. We ght of sentence can be carr ed out n term s of we ght of paragraph and themat- c conceptual vector space model , after the we ght of all the sentences have been carr ed out, the w e ghts are order ng ac- cord ng to the r magn tude. Sentences w th h gh w e ght are selected as summar zat on sentences . In order to evaluate the summar zat on system, w e use two d fferent methods; one s to compare mach ne abstracts w th manual abstract . The other s to compare the prec s on between d ff erent methods of summar zat on. Key words: computer appl cat on; Ch nese nformat on process ng; HowNet ; automat c text summar zat on ; conceptual vector space model 1 , Internet , , , , : : 2004 - 07 - 27 : 2004 - 12- 02 : ( 6044200