An Experiment Report On The Validity Of Machine Translation Of Passive Sentences In Scientific Texts

With the development of neural network technology,machine translation has ushered in a new wave of development.In the survey of the domestic language service industry,it can be seen that machine translation is more and more widely used in this field.Among the translation service areas involved,scientific texts have accounted for the largest portion.The previous studies show that machine translation has shown some applicability in translating literary texts.But due to the constraints such as the automatic and mechanical processing of translation,machine translation is more applicable to non-literary texts such as scientific texts.And in English scientific texts,the passive sentence is the most common sentence pattern.So when using machine translation to translate scientific texts,the validity of machine translation of passive sentences affects the translation quality of the whole text to a certain extent.Baidu,Google and Youdao,the three online machine translations based on the neural network technology,were selected as the subjects in the research,and 220 passive sentences in scientific texts of the college English teaching materials were selected as the basic corpus to make a comparative analysis,to explore the validity of machine translation of passive sentences in scientific texts.On the evaluation of BLEU（Bilingual Evaluation Understudy）,the obtained Chinese translations were classified according to the translation techniques adopted by the human translation first,and then Corpus Word Parser was used in this experiment to segment the words to calculate the BLEU scores.When counting the linguistic metrological features,Corpus Word Parser and Free CLAWS web tagger were used to do the POS（part-of-speech）tagging of the texts first,and then the processed texts were retrieved in Antconc.When summarizing the sentence patterns of the Chinese translations,the human translation was classified first according to their sentence features,and then this classification was taken as the reference to divide the sentences generated by the three OMTs.When analyzing the experimental results,combined with the data and the different Chinese translations,the machine translation,human translation and the English original texts were compared based on the translation criteria of EST（English for Science and Technology）.The problems in translating passive sentences in scientific texts using machine translation were discussed via the qualitative and quantitative analysis.The study has found that the three OMTs have some similarities in this experiment.The validity of the three OMTs is not as high as that of the human translation.And on the whole,the three OMTs in this experiment prefer to translate English passive sentences into Chinese active sentences,which is similar to the human translation.Moreover,the three OMTs are more limited by the structure of the original sentence and tend to produce a word-for-word translation.And by comparing the Chinese passive sentences in the four translations,it can be seen that the word “被” is used more often in machine translation than that in human translation,and other passive markers are rarely used to express the passive voice in the OMTs’ versions.This shows that the way machines deal with English passive sentences is not flexible enough,and the sentence patterns in the translation are relatively simple.Through the analysis of the experimental data and the translation results,the three OMTs also have their own characteristics.Among the three,Youdao’s translations have the highest similarity with human translation,and the translation of technical terms is the most accurate.In general,the translations of Youdao are more in line with the expression habits of Chinese readers.But the manifestation of function words is also the most prominent,which shows that the translations are a little more cumbersome than the other two.Google’s translations are the most concise of the three.Google can convey information in short sentences,but the accuracy of information is not as good as Youdao in that the translation of technical terms is not as standard as Youdao.In addition,Google’s translations are closer to the source language in sentence structure,which shows that Google is more inclined to translate according to the word order of the original text.The performance of Baidu in this experiment is acceptable,but it has the problem of mistranslation,which also leads to the fact that Baidu’s translations are not as good as the other two in terms of integrity.Therefore,on the whole,Youdao performs best in this experiment,and it has a good validity in translating passive sentences in scientific texts.Based on the experimental data,this thesis not only enriches the study of passive sentences in scientific texts and puts forward the expectations for the future development of machine translation,but also provides some reference and inspiration for the translator’s post-editing.