人类语言中交叉与距离关系的计量分析  

A Quantitative Study of the Relationship between Crossing and Distance of Human Language

作　　者：陆前 刘海涛[2] 

机构地区：[1]广西师范大学文学院,广西桂林541004 [2]浙江大学外国语语言文化与国际交流学院,浙江杭州310058

出　　处：《山西大学学报（哲学社会科学版）》2016年第4期49-56,共8页Journal of Shanxi University(Philosophy and Social Science Edition)

基　　金：国家社会科学基金重大项目"现代汉语计量语言学研究"(11&ZD188);中国博士后科学基金资助项目(2015M571852)

摘　　要：已有研究从经验数据出发认为人类语言偏爱依存关系不交叉的结构。文章提出一种枚举依存树的方法进行依存句法结构计量研究,结果显示:语言中的依存交叉并非偶然现象,根据计数公式推导出的交叉依存比例远高于真实语言树库的统计数据;平均依存距离难以区分交叉与不交叉依存;依存距离最小化能约束交叉依存比例降至真实语言情况,但此时平均依存距离与真实语言不同。这些说明除受人类认知机制约束外,可能还存在其他因素促使人类语言倾向于不交叉。By analyzing sentence structure empirically,existent researches have revealed that human language prefers the structure of non- crossing dependencies. In this paper,we develop a method of encoding and enumerating dependency tree,which is used for the quantitative research on dependency syntactic structure. The experiment results show： dependency crossings are non- random phenomena; the proportion of crossing dependencies is much higher than statistics results of the treebanks of real languages; mean dependence distance alone can be used as a metric of analyzing the processing difficulty of dependency structure,but it hardly distinguishes between crossing dependencies and non- crossing dependencies; although the lower dependency distance can restrain the proportion of crossing dependencies to that of real language,the average value of dependency distance differs from that of real language. These findings suggest that in addition to human working memory capacity constraints,there may be other mechanisms that make the syntactic structure of human language have a tendency without crossing dependencies.

关 键 词：依存树 依存距离 交叉依存 句法复杂性 依存距离最小化 

分 类 号：H087[语言文字—语言学]