论文部分内容阅读
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。
Correct analysis of non-projection dependence is the bottleneck that restricts the precision of the dependent syntactic analysis. There is also a widespread non-projection dependence in Chinese. By investigating the Pennsylvania Chinese treebank, we can understand the distribution of cross-dependencies in Chinese as a non-projective dependency. If we include the side crossings caused by clause sharing and box prepositions, there are 2849 cross-references, accounting for 16.42% of 15162 sentences of the entire corpus. If limited to the clause range, 1712 sentence cross-dependent phenomenon, accounting for about 11.29% of the total corpus. Remove the side-by-side relationship caused by the juxtaposition, the remaining phenomena can be divided into two major categories of nine subcategories of cross-dependent use.