Generation of natural language by generalized phrase structure grammars (Q4023951)
From MaRDI portal
!
This is the item page for this Wikibase entity, intended for internal use and editing purposes. Please use the normal view instead:
scientific article; zbMATH DE number 108497
| Language | Label | Description | Also known as |
|---|---|---|---|
| default for all languages | No label defined |
||
| English | Generation of natural language by generalized phrase structure grammars |
scientific article; zbMATH DE number 108497 |
Statements
23 January 1993
0 references
natural language generation
0 references
generalized phrase structure grammar
0 references
0 references
0.8932303
0 references
0.8928937
0 references
0 references
0.8715604
0 references
Generation of natural language by generalized phrase structure grammars (English)
0 references
The book provides a thorough treatment of the generation of natural languages based on the formalism of Generalized Phrase Structure Grammar (GPSG). It combines linguistic evidence and computational arguments on an equal level. First, a survey of the field of natural language generation is given, and alternative proposals for natural language generation developed in the area of computational linguistics are considered. Next, a comprehensive description of the theoretical and formal framework underlying GPSG is provided together with a discussion of linguistic problems associated with that approach. Emphasis is laid on a proper consideration of feature instantiation principles. Finally, a constructive version of GPSG is supplied and its suitability of natural language processing assessed. After demonstrating the computational inefficiency of the original formalism, a refined definition of GPSG is given. It forms the basis for a GPSG system that can be used bidirectionally for pasing as well as generation, based on a strict separation of linguistic and control knowledge. Its capabilities are discussed using fragments of English and German language as empirical evidence. Efficient processing strategies for natural language generation within this redefined framework are worked out and the implications for the representation of lexical items in the lexicon and lexicon access are dealt with. Finally, the given implementation is compared with alternative implementations of GPSG systems. Furthermore, two alternative inputs for the generation process are distinguished. Within the grammar- driven approach, syntactic representation structures (encoding stems, number and tense information) are supplied, while the structure-driven approach provides functor-argument structures as input for the generation process. The latter structures particularly reflect the application environment of the generator, viz. a machine translation system, where functor-argument structures represent the transfer level for the translation process. The final chapter contains the descriptions of the generation procedures which are illustrated on the basis of annotated generator runs. In particular, relations between the pattern-action rules used in the structure-driven approach and the well-known production rule system paradigm are discussed together with conflict resolution strategies for multiply applicable rules. The appendix lists the complete GPSG specification as well as pattern-action rules for the German language fragment covered by the implemented version of the generator.
0 references