A Chinese Dependency Syntax for Treebanking

A Chinese Dependency Syntax for Treebanking Haitao Liu, Wei Huang Department of Applied Linguistics Communication University of China CN-100024, Beijing, China {byliuhaitao, howard}@cuc.edu.cn Abstract. This paper presents a Chinese dependency syntax for treebanking. The syntax contains 13 word classes and 34 dependency types. A format of treebank based on the syntax is also proposed for the applications of computational and general linguistic research. Some experiments show that the treebank based on the proposed dependency syntax can be used for training and evaluating the dependency parser and for quantitative analysis of Chinese syntax. Keywords: dependency relation, dependency syntax, treebank, Chinese syntax 1 Introduction Treebanks are often used as a tool and resource for training and evaluating a syntactic parser in compu- tational linguistics [1]. However, treebanks are not only useful to computational linguists, they are also an important tool for linguists from other branches of linguistics. In this way, if we are planning to build a treebank, perhaps we have to consider its applications in these two fields. Although the Penn treebank based on phrase structure is still the standard of computational applica- tions of treebank, many new projects, particularly in Europe, like to use dependency structure as the an- notation schemes [5]. In this paper, we present a Chinese dependency syntax for treebanking. A format of dependency treebank is proposed for training and evaluate a dependency parser and for quantitative analysis of Chi- nese. In section 2, we discuss some kernels of dependency grammar and how to build a dependency gram- mar for a language. Section 3 presents the proposed Chinese dependency syntax and how to process co- ordinating structures in this syntax. In section 4, a format of dependency treebank is given and some questions on the treebanking are discussed. 2 Dependency relation and dependency syntax Sentence analysis based on the dependency relations has a longer history than the method based on phrase structure. The ideas of dependency analysis are found more or less in traditional grammar of many languages. In other words, the school grammars in many countries are similar with the syntactic 126 principles based on dependency relations. This is an important argument, if we consider that building of treebank is a labor-intensive task. What’s dependency relation, this is not a simple question, because “It is perhaps surprising that one question which, as far as I know, has never been seriously addressed in the Dependency Grammar (DG) literature is what the dependency relation might really amount to” [6]. Considering that Kreps ignores many documents in German and French, which are main publication languages of dependency grammar, we cannot agree completely with him, but he reveals a fact that linguists have still different definitions about this concept. Although the definition is not unified, but the following properties are generally accepted by linguists to constitute the core properties of a syntactic dependency relation [4][8][10][11][13]: − It is a binary relation between two linguistic units. − It is usually asymmetrical, with one of the two units acting as the governor …

Content truncated to 500 words