并列复句的自动识别初探  

Automatic recognizing of complex sentences with coordinating relation

作　　者：周文翠[1] 袁春风[1] 

机构地区：[1]南京大学计算机软件新技术国家重点实验室,南京210093

出　　处：《计算机应用研究》2008年第3期764-766,共3页Application Research of Computers

基　　金：国家自然科学基金资助项目(60673043)

摘　　要：从《人民日报》语料库中抽取4万多个句子作为训练集和测试集,选取其主语、谓语等相关特征并根据《知网》将特征量化,然后使用支持向量机进行训练,获取判别并列复句的模型。在开放测试中获得了84%的准确率。This paper extracted nearly 40 000 sentences as training data and testing data from People＇ s Daily newspaper. And then chose some main features such as subject, object and used Hownet to quantify these features. Finally trained these data with support vector machine （SVM） to get the model. The proposed method had about 84 % precision on the test data. The result of experiment indicates that this approach is feasible.

关 键 词：并列复句 知网 支持向量机 

分 类 号：TP301[自动化与计算机技术—计算机系统结构]