An efficient LR parser generator for tree-adjoining grammars

In this chapter we discuss practical LR-like parser generator models for Tree Adjoining Grammars (TAGs) and propose a new algorithm. The algorithm has been implemented and applied to two large coverage TAGs for English: the XTAG English grammar, and a grammar automatically extracted from the Penn Treebank. The generated tables have very favorable characteristics compared to an existing approach by Nederhof, undermining earlier beliefs that LR parsing for TAGs would be inadequate for parsing natural language. Indeed, our parser generator has been used to build fast accurate best-parse parsers for natural language, as reported in (Prolo, 2002a).