Abstract

Post-edit is defined as the way of correcting machine translation (MT) output. The massive use of (MT) systems in the modern world, specifically neural machine translation (NMT) makes post-edit as one of the main skills translators should have to involve in global markets. The present study investigates the improvement of the quality of the product of translation students to post-edited Google MT output. It focuses on the amounts of post-edit in correlation with types of errors produced by Google MT. Moreover, the study investigates the quality of the final products 44 translation students at University of Basrah in their final year are involved in the study. They receive pre-translated text by Google translate and are asked to post-edit. The results have shown that (21.2%) terminology errors and (48.8%) grammatical errors have not been corrected. A mixed-method approach is used to collect qualitative and quantitative data analyzed within the Dynamic Quality Framework (DQF) adapted by Translation Automation User Society (TAUS). TAUS error typology is used as a model to assess participants’ outputs and to be evaluated by a jury of professional instructors. The findings have shown that there is an improvement in quality due to the statistical analysis of the quantitative data which have shown a significant correlation between post-edit practice (PEP) and post-edit quality (PEQ) as p value = 0.003. The statistical results imply an improvement of post-edited output quality.

Keywords

The article was added to IASJ on 2021-10-20

418 Total full text downloads since the date of addition