Relational-grammar-based generation in the JETS Japanese-English machine translation system

Abstract

This paper describes the design and functioning of the English generation phase in JETS, a limited transfer, Japanese-English machine translation system that is loosely based on the linguistic framework of relational grammar. To facilitate the development of relational-grammar-based generators, we have built an NL-and-application-independent generator shell and relational grammar rule-writing language. The implemented generator, GENIE, maps abstract canonical structures, representing the basic predicate-argument structures of sentences, into well-formed English sentences via a two-stage plan-and-execute design. This modularity permits the independent development of a very general, deterministic execution grammar that is driven by a set of planning rules sensitive to lexical, syntactic and stylistic constraints. Processing in GENIE is category-driven, i.e., grammatical rules are distributed over a part-of-speech hierarchy and, using an inheritance mechanism, are invoked only ff appropriate for the category being processed.