Dependency Unification Grammar

DEPENDENCY UNIFICATION GRAMMAR Peter Hellwig University of Heidelberg, D-6900 Heidelberg, West Germany Abstract This paper describes the analysis component of the language processing system PLAIN from the viewpoint of unification grammars. The principles of Dependency Unification Grammar (DUG) are discussed. The computer language DRL (Dependency Representation Language) is introduced J.n which DUGs can be formulated. A unifi- cation-based parsing procedure is part of the formal- ism. PLAIN is implemented at the universities of Hei- delberg, Bonn, Flensburg, Kiel, Zurich and Cambridge U.K. 1. Introduction The recent development of grammar theory e×hibits convergencies among various approaches, such as Gov- ernment-Binding Theory, Generalized Phrase Structure Grammar, Definite Clause Grammar, Lexical Functional Grammar, Functional Unification G~ammar, and others. To varying degrees these theories share the following principles: (i) They take into account dependency rela- tions, using notions such as "head" or "governor". Phenomena such as long distance dependencies are viewed as tou¢:hstones for the formalisms. (ii) They pay attention to functional aspects. The representation of syntactic roles is seen to be a task. (iii) They agree that syntax must be lexically restricted and thus place a large portion of the grammatical information in the lexicon. (iv) They base their algorithms on the princi- ple of unification, i.e. complex categories are brought into agreement in the syntactic context. These common features make it possible to com- pare the solutions of the different formalisms as well as their problems. The main difficulty for the computer application of unification grammars lies in their complexity. LFG, for example, bases its syntactic c-structures on the phrase structure principle, while the functional f-structures represent dependency relationships between functors and a~guments. This causes problems for the parser which needs information on both structures while it is creating them. The development of GPSG seems to be marked by the effort to introduce more and more types of rules so as to adequately constrain the formalism. As a result, the control of analysis is distributed over many resources and is, therefore, increasingly difficult. Since a large number of constraints are of a lexical nature, the lexicon becomes more and more unwieldy in all of the formalisms. Common advantages and common problems of unifi- cation grammars suggest examining strategies from still other frameworks. This is to be done here with respect to dependency grammar. DUG rearranges the available means of description. As a result, the be- nefits of the common principles are fully felt where- as the difficulties mentioned are largely avoided. 2. Dependency Representation Language (DRL) Grammar formalisms and computer languages are usually developed independently. DRL is both at the same time. In the same spirit as PROLOG is tailor-made for the purposes of logic, DRL has been particularly adapted to represent linguistic structures. Whereas the interpreter for PROLOG includes a theorem prover, the interpreter for DRL is linked with a parser. (DRL also serves for the purpose of knowledge representa- tion within the deduction component of PLAIN. This aspect will not be discussed here.) DRL consists of bracketed expressions which are …

Content truncated to 500 words