The Linguistic Relevance of Tree Adjoining Grammar
AUTHOR: Anthony Kroch and Aravind Joshi
DATE: June 1985
PUBLICATION: CIS Technical Report MS-CIS-85-16
EMAIL ADDRESS: joshi@linc.cis.upenn.edu, kroch@change.ling.upenn.edu
PDF FILE NAME: relevance3.pdf
ABSTRACT:
In this paper we apply a new notation for the writing of natural language grammars to some classical problems in the description of English.
The formalism is the Tree Adjoining Grammar (TAG) of Joshi, Levy and Takahashi 1975, which was studied initially only for its mathematical properties but which now turns out to be an interesting candidate for the proper notation of meta-grammar; that is for the universal grammar of contemporary linguistics.
Interest in the application of the TAG formalism to the writing of natural language grammars arises out of recent work on the possibility of writing grammars for natural languages in a metatheory of restricted generative capacity (for example, Gazdar 1982 and Gazdar et al. 1985).
There have been also several recent attempts to examine the linguistic metatheory of restricted grammatical formalisms, in particular, context-free grammars.
The inadequacies of context-free grammars have been discussed both from the point of view of strong generative capacity (Bresnan et al. 1982) and weak generative capacity (Shieber 1984, Postal and Langendoen 1984, Higginbotham 1984, the empirical claims of the last two having been disputed by Pullum (Pullum 1984)).
At this point the TAG formalism becomes interesting because while it is more powerful than context-free grammar, it is only "mildly" so.
This extra power of TAG is a direct corollary of the way TAG factors recursion and dependencies, and it can provide reasonable structural descriptions for constructions like Dutch verb raising where context-free grammar apparently fails.
These properties of TAG and some of its mathematical properties were discussed by Joshi 1983.
It is our hope that the presentation will support the claim, currently controversial, that the exploration of restrictive mathematical formalisms as metalanguages for natural language grammars can produce results of value in empirical linguistics.