Essentials of Arabic Machine Translation Evaluation: The Potentials and Limitations

Evaluation remains an open fundamental issue and one of the most important stages in the lifecycle of an MT system. With the rapid growth in sales of MT software and the increasing availability of MT services over networks, there is an urgent need for MT researchers, developers and vendors to agree and implement objective , reliable and publicly acceptable standards of MTE. The book therefore investigates the overall quality of three currently available English-into-Arabic MT systems. The evaluation of the cognitive , linguistic and computational capabilities of an MT system is clearly a pivotal issue. The theoretical construct adopted in this study is based on the Framework for Evaluation of MT in ISLE (FEMTI, 2003). The evaluation focuses on selected functional and non- functional criteria of system's performance. The appropriate metrics, scoring scheme and methods have been identified and described. The findings should assist systems' designers and developers to find ways of improvement. Further, they should be useful to potential users to determine which of these systems can best satisfy their needs.