Machine Translation Quality Assessment Based on Entropy Weight TOPSIS Method

Translation quality assessment has become a hot topic in various fields, but the traditional qualitative analysis cannot meet the needs of diversified translation criteria.Therefore, it is necessary to screen different quality assessment indicators and establish an evaluated model to meet the comprehensive requirements.Three first-level indicators: fluency, accuracy, logicality, ten second-level indicators such as mistranslation, over-translation, and missing translation, and nine third-level indicators such as mistranslations in parts of speech and acronyms are selected to integrate the assessment criteria of words, sentences, and discourses into the framework of a multi-dimensional quality metrics (MQM), so as to build a more detailed and hierarchical translation quality assessment model.Based on the original features of the data, the entropy weight method is adopted to calculate the index weight, and the TOPSIS method is adopted to calculate the ranking of the mainstream neural network machine translation.Machine translation can optimize their systems based on the real-time feedback of post-editing.Human-computer interaction can be achieved by integrating the latest theories and research findings of linguistics, cognitive science, and natural language processing.