Types and Distribution of the Dependence of Arc Covering the Top Node in Mandarin Chinese
摘要
The correct handling of non-projective dependency is very important for improving the accuracy of dependency parsing. As a kind of non-projective dependency phenomenon, the dependency of arc covering the top node is also widely existed in Mandarin Chinese, and has different characteristics. By investigating the corpus of U-Penn Chinese Treebank (CTB 5.0), we know about the distribution of this phenomenon in Mandarin Chinese. The 18782 sentences in the corpus can be decomposed into 52853 clauses, of which 231 clauses have this phenomenon, accounting for 0.44%. These phenomena can be divided into two categories and six subcategories. Sorting out the dependency of arc covering the top node is of reference significance to the syntactic parsing and linguistic research of non-projective dependence.