汉构:面向深层语言处理的语法工程

ManGO: Grammar Engineering for Deep Linguistic Processing

【目的】开发面向深层语言处理的汉语普通话在线语法(简称汉构)。【应用背景】汉构是在DELPH-IN环境内,基于语法母体,在LKB平台上开发的可计算汉语语法。它的句法和语义分析的理论框架分别是中心语驱动的短语结构语法和最简递归语义。汉构为进一步开发资源型语法和商用奠定良好基础。【方法】根据系统的语言学本体研究对语言知识进行形式化描写;汉构的计算实现经历语法定制、汉语MRS测试套件、词库建设、语法规则定义和MRS描写等环节。【结果】汉构覆盖汉语基本词类和主要语言现象,完全覆盖MRS测试套件。【结论】汉构是最早的中型可计算汉语语法之一,是形式语法理论和计算语言学领域间开展合作研究的桥梁和有效载体。

[Objective] This article contributes to the development of ManGO （Mandarin Grammar Online） for deep linguistic processing. [Context] On the platform of LKB （the Linguistic Knowledge Builder） and based on Grammar Matrix, ManGO is developed in the environment of DELPH-IN （Deep Linguistic Processing with HPSG Initiative）. The frameworks of its syntactic and semantic analysis are HPSG （Head-driven Phrase Structure Grammar） and MRS （Minimal Recursion Semantics） respectively. ManGO lays a solid foundation for further resource grammar development and commercial application. [Methods] First, linguistic knowledge is formalized according to systematic Ontological studies. Then, the computational implementation of ManGO goes through grammar customization, creation of a Chinese MRS test suite, lexicon building, definition of grammar rules and MRS representation. [Results] ManGO covers nearly all the major Chinese word types and grammar phenomona, and fully covers the Chinese MRS test suite. [Conclusions] ManGO is one of the earliest medium-size computational grammars of Chinese. It serves as the bridge and effective carrier of the interdisciplinary studies across formal grammar theory and computational linguistics.