Junction grammar and machine assisted translation

TLDR: The Junction Grammar model of language is examined, then the model of translation suggested by themodel of language, and finally an application of the model to mechanical translation or, as the authors prefer to term it, computer-assisted translation are examined.

Abstract: 1. The Junction Grammar Model of Language. Junction Grammar is a generative grammar with multiple levels of representation. The deep level is a nonlexical conceptual level, where the meaning of a segment of discourse is represented by a set of intersecting branching trees, each tree consisting of conceptual units (or sememes) and junctions. 1 The junction patterns (whence the name, Junction Grammar) are selected from a finite language-independent pool, and are used to describe the relationships between the sememes. The basic concepts of Junction Grammar were developed at Brigham Young University by Dr. Eldon G. Lytle from 1968 to 1970, but significant refinements to the theory continue to be added. We will first examine the Junction Grammar model of language, then the model of translation suggested by the model of language, and finally an application of the model to mechanical translation or, as we prefer to term it, computer-assisted translation. The Junction Grammar model distinguishes four levels of representation (see Fig. 1). Level I is the real world, where objects represent themselves. Level II is the conceptual level at which we postulate that a person has a pool of base sememes from which he draws to form segments of discourse by the application of junction rules. These sememes correspond to a particular meaning of a word, phrase, or part of a word at level IV. At level II, we call a meaningful segment of discourse a “Well-Formed Syntacto-semantic Statement” or WFSS (pronounced /wufəsəs/). A WFSS is a statement about the syntax which joins a set of concepts to form a segment of discourse. This level II syntax is not to be confused with surface syntax, which deals with such matters as word order. Our 1 Other work in the area of conceptual grammar includes that of Roger Schank at