The Fulcrum Approach to Machine Translation
Christine A. Montgomery · 1997
problems “in terms of the two components of the term: machine problems, and translation problems.” While the machine problems made us crazy, the translation problems made us think differently about language than we might otherwise have done, which has had some advantages and some disadvantages in the long run. I will save anecdotes about the former and comments about the latter for the discussion. In this paper I will focus on the translation problems and, in particular, the translation approach that was developed by Paul Garvin, with whom I was associated, initially at Georgetown University, and later in the Synthetic Intelligence Department of the Ramo-Wooldridge Corporation and successor corporations: Thompson Ramo Wooldridge and Bunker-Ramo. The initial MT experiments at Georgetown University in 1954 were designed by Garvin and carried out jointly with IBM, who implemented Garvin's design on an IBM 701 computer. These experiments were necessarily quite limited in scope and were a demonstration of possibility rather than feasibility. The approach Garvin developed subsequently, entitled “The Fulcrum Approach to Machine Translation,” was based on what he referred to as the “fulcra ” of syntactic phrases of various types. The fulcrum of a syntactic unit is the constituent that “provides the best point of leverage for its translation, ” essentially because it governs the syntactic behavior of the unit, functioning as the head of the given syntactic unit