New Approaches to Machine Translation

The current resurgence of interest in machine translation is partially attributable to the emergence of a variety of new paradigms, ranging from better translation aids and improved pre and post-editing methods, to highly interactive approaches and fully automated knowledge-based systems. This paper discusses each basic approach and provides some comparative analysis. It is argued that both interactive and knowledge based systems offer considerable promise to remedy the deficiencies of the earlier, more ad-hoc post-editing approaches, 59 1. A Historical Perspective Researchers in machine translation have aspired for three decades to develop highly-accurate, practically-useful, fully-automated translation systems. This ultimate objective remains as elusive today as it was in the late 1950's, although the field has seen considerable progress ranging from theoretical advances in computational linguistics to useful partially-automated translation systems. In the early heyday of machine translation, the rallying cry was "95 % accurate, fully automatic high quality translation! " [13, 2]. In fact, that motto was repeated so often than it became an acronym