The Translation Quality Assessment of Mainstream Neural Machine Translation Tools on Multidimensional Quality Metrics

Compared to Statistical Machine Translation, Neural Machine Translation significantly improves translation quality by leveraging deep learning models to handle long-range dependencies and complex sentence structures. However, traditional machine translation quality evaluation methods, such as Bilingual Evaluation Understudy and Translation Edit Rate, rely on lexical matching with reference translations andfail tofully capture more nuanced qualityfeatures such as grammar, context, andfluency. Within this context, this study evaluates the translation quality of mainstream neural machine translation tools- Youdao Translate and Google Translate - on information-oriented texts using Multidimensional Quality Metrics and the MTMARK tool. The study aims to promote the optimization of translation technology, improve the scientificity and objectivity of translation quality assessment, and provide an important reference for translation education andpractice.