| With rapid development of machine translation,various smart translation platforms become accessible to the public for general translation or translation for specialized fields.A number of previous studies have discussed the quality of machine translation output for general machine translation.However,few studies were conducted on evaluating the quality of machine translation for specified fields.This study compares the translation quality of general machine translation system and specialized machine translation system.The research selected Smart Translation System of Standardization(the STSS platform)which self-developed by China National Institute of Standardization(CNIS)as the specialized machine translation platform,then selected Google Translate and Youdao Translation as the general machine translation platform.And the study compiled six documents of international standards(English as the source language,in total 9639 words)and six documents of Chinese standards(Chinese as the source language,in total 12611 characters)as source texts.Using BLEU(Bilingual Evaluation Understudy)as the main measurement,we then compared the quality of machine translation for the ten documents by Google Translate,Youdao translation and the STSS platform.Our results show that the STSS platform,which has relatively larger specified translation memory for the field of international and Chinese standards,performed significantly better than Google Translate and Youdao translation both in Chinese to English and English to Chinese translation when the source text included in the translation memory of the STSS platform.However,the score of the three machine translation platforms did not show significant difference when the source text is not included in translation memory of the STSS platform.We then manually compared the translation errors based on the machine translated target texts based on the Quality Evaluation Code for Localization Translation and Desktop Publishing issued by Translators Association of China.Accordingly,from the perspective of the number of error types,the error types with the highest proportion are "terminological errors" when translating standard documents.From the perspective of error categories and levels,both specialized machine translation system and general machine translation system have serious errors which can affect the readers’understanding of the original text.This pilot study summarized the effectiveness of a self-developed machine translation platform for a specified field and translation errors in detail,providing a perspective on evaluating the translation quality of a smart domain-specific translation platform.It also provided useful implications for educators,researchers,and designers of smart machine translation platforms. |
