摘要
非投射性依存现象的正确分析是当下制约依存句法分析精确率的瓶颈问题,汉语中也广泛存在非投射性依存现象。通过调查宾州汉语树库,我们可以了解作为非投射性依存现象之一的交叉依存现象在汉语中的分布情况。如果包括因分句共享与框式介词等引起的边交叉,交叉依存有2849句,约占整个语料库15162句的16.42%。如果限定在分句范围内,1712句有交叉依存现象,约占总语料的11.29%。除去因并列关系引起的边交叉,其余现象可以分为两大类共九个小类的交叉依存使用情况。
The parsing of non-projective dependency has been the bottleneck problem for the accuracy of dependency parsing system. There is extensive existence of non-projective dependency in Mandarin Chinese. Based on the U-penn Chinese Treebank,the paper investigates the distributions of edge-crossing dependency,an example of nonprojective dependency, in Mandarin Chinese. It is found that when edge crossings produced by clause sharing and circumposition are included,edge-crossing dependency can be found in 2849 sentences,about 16. 42 percent of the 15162-sentence corpus; when the range is limited to clauses,1712 sentences are identified,accounting for 11. 29 percent of the overall sentences; and apart from edge crossings caused by coordination,other usages can be divided into two categories with nine sub-categories.
出处
《语言文字应用》
CSSCI
北大核心
2015年第1期90-97,共8页
Applied Linguistics
关键词
依存
交叉依存
非投射性依存
分布
dependency
edge-crossing dependency
non-projective dependency
distribution
