Machine Automatic Translation Evaluation Based on Big Data Algorithms

Currently,research on machine translation is mainly divided into four stages: dictionary matching method, comprehensive linguistic knowledge method, corpus based statistical machine translation, and neural machine translation. Machine translation evaluation is a hot topic in current machine translation research, which can effectively identify defects in translations and improve translation quality. The evaluation of translation quality can be divided into two types: manual evaluation and computer evaluation. Manual evaluation has higher accuracy due to being completed in the early stages, but manual evaluation requires a lot of manpower. In addition, human evaluation has a certain degree of subjectivity. The rise of automated evaluation technology has made up for the shortcomings of traditional manual evaluation methods. The BP (back propagation) algorithm discussed in this article can quickly evaluate translations and has higher efficiency compared to manual evaluation. At present, most machine translation evaluation methods evaluate the quality of a translation based on the similarity between the translation and the reference translation. Therefore, in the evaluation process, the amount of translation materials cited by the translator and the utilization rate of the obtained materials are the key factors determining the effectiveness of the evaluation method. This article used the BP algorithm to overcome the bottleneck problem of overly relying on sample data, extract feature vectors of translations well, and improve the accuracy of translation evaluation. Finally, this article verified through experiments that the machine translation automatic evaluation model based on deep learning algorithms was lower than other models, and its error rate remained below 0.24.