چکیده:
Many studies abroad have focused on neural machine translation and almost all concluded that this method was much closer to humanistic translation than machine translation. Therefore, this paper aimed at investigating whether neural machine translation was more acceptable in English-Persian translation in comparison with machine translation. Hence, two types of text were chosen to be translated by Google Translate. The inputs have been translated in two distinctive methods. The outputs were investigated by the descriptive-comparative human analysis model of Keshavarz. Consequently, the results revealed that approximately the same errors were found in both methods. However, semantic aspects were improved.
خلاصه ماشینی:
A Comparative Study of English-Persian Translation of Neural Google Translation Mina Zand Rahimi Shahid Bahonar University of Kerman mina_zandrahimi@uk.
In other words, “MT involves the use of computer program to translate texts from one natural language into another automatically,” said Ping (as cited in Baker & Saladanha, 2009, p.
Neural Machine Translation In an end-to-end approach of NMT, there was a significant ability of learning the way of connecting input elements to output’s.
1. Materials The materials considered in this paper included English texts as the input and Persian text as the output data.
Procedures A descriptive-comparative human analysis model of Keshavarz (1999) were used in the paper by which the translated texts were analyzed and each resulted error were comparatively observed.
These collected data (both input and output) included about 1350 words and among which the percentages of each error has been provided.
Frequency Percentage of Errors Error Type GMNT (%) GMT (%) Word disordering 63% 67% Verb-mismatching 3% 4% Wrong collocation 20% 13% Wrong preposition 2% 3% Active and passive 2% 0% Tense error 0% 0% Mistranslated* 3% 6% Not translated 2% 13% *Mistranslated error has been considered linguistically and not semantically text.
4. Conclusion According to what has been analyzed, the result of the research revealed that GMNT, in comparison to GMT, was more successful in semantic aspects, especially in Persian-to-English translation.
Moreover, based on the percentage of the error reduction it was concluded that GNMT system was beneficent for Persian translation in Google Translate.