一种基于词汇功能描述的树库构建方法

Constructing Tree-Bank Based on Lexical Functional Description

提出在面向数据的英汉机译系统中 ,一种自动构建包含丰富语法与句法信息的树库的方法·即 ,基于词汇功能描述 ,将源语语料库中各句子转换为组成结构与功能结构的组合表示形式·其中 ,组成结构表示单词至较大单元或短语成分的分层组成 ,以一种比较简单的短语结构树的形式表示·而功能结构为语法功能如主语 ,宾语等等的表示·从目前的实验结果来看 ,已经达到比较令人满意的自动标注效果·因此 ,基于树库的成功构建 ,可有效且准确地进行源语分析·

A method of constructing tree bank is presented that contains rich grammatical and syntactical information in a DOP based English Chinese machine translation system. In the implementation, according to lexical functional description, every sentence in the source language corpus was converted into a combined representation of a constituent structure tree and a functional structure. The constituent structure was designed in a kind of hierarchical structure from words to units or phrases, which is shown in a simply form of phrase structure tree. The functional structure was formed to play the syntactical functions, such as subject, object and so on. It can be seen that the automatically tagging effect is very satisfactory by limited experiments. So based on the successful construction of tree bank, the source language analysis can be completed efficiently and accurately.