Lexical Transfer: A Missing Element in Linguistics Theories

One of the necessary tasks of a machine translation system is lexical transfer. In some cases there is a one-to-one mapping from source language word to target language word. What theoretical model is followed when there is a one-to-many mapping? Unfortunately, none of the linguistic models that have been used in machine translation include a lexical transfer component. In the absence of a theoretical model, this paper will suggest a new way to test lexical transfer systems. This test is being applied to an MT system under development. One possible conclusion may be that further effort should be expended developing models of lexical transfer. i. An Early Approach to Lexical Transfer Years before the machine translation community was burdened with guilt by the ALPAC report, David Hays, former chairman of the International Committee on Computational Linguistics, proposed a procedure for lexical transfer (Hays, 1963, pp 205-208). We will describe it, quoting pieces to preserve the original flavor. i. a. File of Equivalent-Choice Data "Most words...have uniform translations, but not all." "These exceptions to the general rule must be discovered and taken into account. The procedure is simple and straightforward. A file of equivalent-choice data...is required." This file is prepared usina real text. When a word is encountered for the first time, one translation is selected and entered into a bilingual glossary. When the same word is encountered again, the human translator/editor attempts to use the translation already in the glossary. Additional translations are added only when the one(s) in the glossary are not acceptable. This procedure is supposed to avoid entering interchangeable alternatives that are only stylistic variations. The file of equivalent-choice data mentioned above is a record of how many times each translation was used.