Introduction

An important characteristic of an FB-LTAG is that it is lexicalized, i.e., each lexical item is anchored to a tree structure that encodes subcategorization information. Trees with the same canonical subcategorizations are grouped into tree families. The reuse of tree substructures, such as wh-movement, in many different trees creates redundancy, which poses a problem for grammar development and maintenance [#!vijay-schabes92!#]. To consistently implement a change in some general aspect of the design of the grammar, all the relevant trees currently must be inspected and edited. Vijay Shanker and Schabes suggested the use of hierarchical organization and of tree descriptions to specify substructures that would be present in several elementary trees of a grammar. Since then, in addition to ourselves, Becker, [#!becker94!#], Evans et al. [#!Evans95!#], and Candito[#!Candito96!#] have developed systems for organizing trees of a TAG which could be used for developing and maintaining grammars. Our system is based on the ideas expressed in Vijay-Shanker and Schabes, [#!vijay-schabes92!#], to use partial-tree descriptions in specifying a grammar by separately defining pieces of tree structures to encode independent syntactic principles. Various individual specifications are then combined to form the elementary trees of the grammar. The chapter begins with a description of our grammar development system, and its implementation. We will then show the main results of using this tool to generate the Penn English grammar as well as a Chinese TAG. We describe the significant properties of both grammars, pointing out the major differences between them, and the methods by which our system is informed about these language-specific properties. The chapter ends with the conclusion and future work.