Machine Translation: The Case of Arabic- English Translation of News Texts

Abstract

This study aims at identifying the common types of errors in Google Translate (GT) in the translation of informative news texts from Arabic to English, to measure the translation errors quality and to assess the fluency and the semantic adequacy of the translation output, and therefore to explain the extent a human translator is needed to rectify the output translation. For this purpose, some examples were purposively selected from online newspapers. The collected data was analyzed using a mixed method approach, as the errors were qualitatively identified, guided by Hsu’s (2014) classification of machine translation errors. Quantitative descriptive approach was used to measure the translation errors quality, using the Multidimensional Quality Metrics and Localization Quality Evaluation. As for assessing the semantic adequacy and fluency, a questionnaire that was adapted from Dorr, Snover, and Madnani (2011) was used. The results of the analysis show that omission, which is a lexical error and inappropriate lexical choice, which is a semantic error are the most common errors. Inappropriate lexical choice is sometimes a result of the homophonic nature of some source text words which can be misinterpreted by the machine translation system. This study concludes that it is useful to use machine translation systems to expedite the translation process, but that accuracy is sacrificed for the sake of ease (less work for the human) and speed of translation. If greater accuracy is required, or desired, a human translator must at least proofread and work on the material.