همایش رویکردهای میان رشته ای به آموزش زبان، ادبیات و مطالعات ترجمه

صفحه اصلی / پنجمین همایش رویکردهای میان رشته ای به آموزش زبان، ادبیات و مطالعات ترجمه

Investigating Syntactic Errors Made by Two Neural Machine Translation Systems, Google Translate and MateCAT, in the Psychology and Medical Domains: A Corpus-based study

نویسندگان :

Abolfazl Khorasani Zadeh Gazki¹ Zahra Amirian²

1- دانشگاه اصفهان 2- دانشگاه اصفهان

کلمات کلیدی :

Neural Machine Translation،Google Translate،MateCAT،Syntactic Relations،Transfer

چکیده :

This study compared syntactic errors in Google Translate with MateCAT. This study used medical and psychology texts. Psychology domain texts were randomly selected from Kaplan and Sadok's Synopsis of Psychiatric, and medical domain texts from the Guyton and Hall Textbook of Medical Physiology. The study identified erroneous punctuation, adverb and adjective misuse, object placement, pronoun use, preposition and postposition placement and usage, verb and subject agreement, verb misplacement, and adverb and adjective misuse. Both MT systems had these errors across domains. Google Translate translated 50.6% of psychology and 49.5% of medicine accurately. MateCAT has a 50.9% perfect translation rate in psychology and medicine. The two MT systems translated identically in all categories except verb and subject agreement in medical data, according to McNemar tests.