Machine Translation and Automatic Language Data Processing

This chapter discusses machine or automatic translation of natural languages. It reviews the status of the art at present; explains its basic operations, methods and procedures; indicates its objectives and uses, and situates machine translation or MT in the general field of automatic language data processing. Finally, it suggests its possible role in language communication as a whole. Machine translation is a relatively new area of automatic language data processing. It came about in part as a result of the conjuncture of three trends: (1) the development of structuralist procedures in linguistics; (2) the increasing sophistication of programming techniques, and (3) the growing capabilities and versatility of computation devices. It also became a subject of interest in the scientific and managerial communities as a result of the increasing volume and diversity of scientific and technical writings in the several languages of scientifically creative cultures, and the lengthening lag between the publication of information in a given language and its accessibility in one or several other languages. A decade ago machine translation was of interest to a relatively small group of people coming from such apparently unrelated fields as philosophy, physics, mathematics, sociology, logic, computational engineering, chemistry, and of course linguistics and languages. This diversity of background among the early comers was to bring about a widely diversified and divergent set of notions as to what automatic translation is or should be, what it ought to try to do, how and why it should do it. Notwithstanding these divergences, MT research today is pursued in a number of centers and laboratories in some twenty countries, including besides the United States, where oriented research may be said to have originated, Great Britain, the U.S.S.R., Japan, Italy, France, Belgium, Germany, and others. The first public demonstration of feasibility was carried out jointly by Georgetown University and IBM in January, 1954, on the basis of an experiment for the transfer of a small corpus of Russian