Error Analysis of Neural Machine Translation in Technical Texts: Google Translate as a Case Study

The reliability of neural machine translation has increased in various fields. However, the accuracy of machine translation is still
questionable, even with advancements in neural machine translation, but the situation should be better when dealing with technical texts that
are known to be clearer, more precise, and fixed. Therefore, the necessity for human intervention in revising the translation should be to a
limited extent. This study aimed to investigate the translation errors faced by neural machine translation, represented by Google Translate when
translating technical texts. It adopts an error analysis approach to evaluate the quality of the aforementioned neural machine translation by
examining its translation of a technical text from Arabic into English and comparing it with a human-certified translation. It also evaluates the
extent of the necessity for human intervention in revising the translation. The results indicate some errors in Google translation, varying from
comprehension and linguistic errors to translation errors, highlighting the necessity for human intervention. Google translation has proven to
be better than human translation in several respects. The implications of this research indicate the remarkable performance of Google Translate
surpassing human translation in several contexts, which can be used in translating technical texts with the need for human intervention.