Machine translation: practical issues

Many of the problems with (and consequently much of the resistance to) machine translation (MT) has sprung from ignorance: on the part of early developers, who failed to appreciate the magnitude of the problems they faced, and hence made outrageous claims; on the part of opponents, who frequently fail to understand what translation is about, and how it is (properly) performed; on the part of zealots for theoretical formalisms, who seldom if ever try to produce working systems and consequently lack appreciation of the magnitude of the practical problems to be faced; on the part of modern developers, too many of whom know little about translators' needs and other human engineering concerns; and on the part of translators, who often fail to realize the potential for improvement in the quality of their working lives when computers become able to take over much of the 'grunt work' in translation. In this chapter we confront these and other problems. We characterize areas where knowledge has been lacking, and discuss the consequences of such ignorance; provide background knowledge enabling the reader to acquire a more balanced perspective from which to judge the developments in the field; and finally present the author's opinions, based upon his experience as an MT system developer, regarding some of the most important issues in the construction of a practical system. This will entail, in places, an exposition of certain aspects of the underlying technology; however, complete understanding of these technical matters is not a prerequisite for appreciation of the major points.