The Treatment of Grammatical Categories and Word Order in Machine Translation

Machine Translation (MT) systems are now in active use around the world. This paper investigates the applicability of current and foreseeable MT technology to translation between one or more of modern Western European languages and Arabic. After an introduction in which we briefly sketch the history and current status of MT and comment on the situation vis-a-vis Arabic translation, we present some general design constraints for state-of-the-art MT systems before proceeding to consider problems posed by the Arabic language in particular. We then outline some approaches to the solutions of such problems, and indicate what special constraints these place on an MT system's design. We next present an architecture for a system that could handle Arabic, and draw some conclusions regarding the prospects for near-term application of such a system. We will not in this paper survey the field of MT; several such treatments are available elsewhere (e.g., [Slocum, 1984]).