A Comparative Analysis of Human Translation and Machine Translation from the Principle of “Faithfulness, Expressiveness and Elegance”从“信达雅”原则对比分析人工翻译和机器翻译开题报告

1. 研究目的与意义（文献综述包含参考文献）

1. introduction1.1 background of the research earlier before, some scholars (hutchins, w. j., faithfulness, expressiveness and elegance proposed by yan fu is universally acknowledged as the optimal translation standard (wang zhenping, 2000). the three criteria indicate that the translation should convey what the source text exactly wants to say, meanwhile be syntactically and semantically correct and, most importantly maintain the style of the original. therefore, in order to compare the differences between human translation and machine translation, it is rational to adopt the principle of faithfulness, expressiveness and elegance. 1.2 purposes of the researchthe research is aimed at achieving three main goals. the first purpose is to find out the differences between human translation and machine translation from the perspective of faithfulness, expressiveness and elegance. secondly, the research will provide evidence that the two types of translation closely relate with each other and complement each others advantages. students, especially english majors, will therefore gain a deeper understanding of the translation skills and techniques they have learned in class and combine them with the application of up-to-date translation tools when practicing. at last, from the long term, the results of this research will enable translators to make a rational choice at work and give some inspirations to the development of machine translation in the future. 1.3 organization of the research this thesis contains five parts in total. chapter one gives a brief introduction of research background, research purposes and the organization of the research. chapter two overviews relevant theory and previous studies to figure out what has been achieved and what has not been achieved. in chapter three, the methodology is presented, including materials, subjects, the design and execution of the research. chapter four reports the findings based on the data collected and discusses the findings in detail. the final part summarizes the major findings, the theoretical and practical implications, reflects on the limitations and gives suggestions for the future research. 2. literature reviewin this information era, the traditional human translation cannot satisfy the increasing need of exchanging large quantities of information within the minimum time, which prompts the development of machine translation. the discussion as to which one does better has never stopped. many researchers who are curious about how machine translation differs from human translation and how well they behave in practice dig into this field. the premise of translation evaluation is that a translation criterion should be set to measure the quality of the translation. the principle of faithfulness, expressiveness and elegance is known as the widely accepted translation standard in china. a great number of studies are aimed at comparing machine translation with human translation from different angles; however,few were done from this principle. even if some researchers made a comparison according to this criterion, the conclusion is often drawn through theoretical analysis. 2.1 machine translation and human translationmachine translation, or automatic translation, refers to the process of converting one human language to another automatically by machines (arnold, d. et al., 1994). the idea of applying computers to translation came out in 1946. the development of machine translation actually corresponds to the call of different times (hutchins, w. j., sinhal, r. a., hu, sun, 2016; liu, 2019). arnold, d. et al. drew a conclusion that machine translation systems have difficulty in solving the problem of ambiguity, lexical and structural mismatches and idioms and collocations. the research conducted by sun jin (2016) showed that machine translation demonstrates higher applicability to informative text than expressive and operative text. 2.2 translation evaluation and the principle of faithfulness, expressiveness and elegancethere are three main approaches to machine translation evaluation, operational evaluation, declarative evaluation and typological evaluation (arnold, d., the second approach makes an evaluation based on certain standard (e.g. liao, 2013; tian, 2018); and the typological evaluation works from the aspect of linguistics (e.g. sun, 2016). in liao menglin (2013)s experiment, the translation is evaluated from the ast (adequacy scale task) standard while tian jiangbo (2018) analyzed the machine translation from the principle of faithfulness, expressiveness and elegance. the declarative evaluation is characterized by low cost, easy operation and derivable results (arnold, d., lee, h., 1993). moreover, yan fus faithfulness, expressiveness and elegance is widely recognized in china. tians research is done through theoretical analysis. this paper will integrate the declarative evaluation based on the principle of faithfulness, expressiveness and elegance with quantitative experiment which has not yet been done in this area. references arnold, d. (2003). why translation is difficult for computers. in h. somers (ed.), computers and translation: a translators guide (pp. 119-142). amsterdam: john benjamins. arnold, d., 以四类文本翻译为例，《湖北师范学院学报（哲学社会科学版）》，33（3）：145-148。

刘天泽（2019），浅谈人工智能翻译和人工翻译的比较与展望，《海外英语》，16：46-47。

穆雷（2006），翻译批评与翻译标准英汉/汉英翻译测试研究系列（三），《外语与外语教学》，4：45-47。

2. 研究的基本内容、问题解决措施及方案

The study is to answer the following questions: 1. What are the differences between human translation and machine translation from the principle of faithfulness, expressiveness and elegance? 2. What inspirations does this research have on translators and machine translation companies nowadays? Research design1. Methods: This research mainly uses the comparative method to analyze the differences between human translation and machine translation. With the fast development of high technology, the contradiction between human translation and machine translation has been heatedly discussed in the translation field. Scholars hold different opinions towards the standard of translation, among which Yan Fus faithfulness, expressiveness and elegance is widely recognized in China. Faithfulness demands the translation convey what the original exactly wants to say; to be expressive is to correctly present the meaning in a more native way; last but not least, the level of elegance which pursues the original writing style and theme is the hardest to reach. In order to know how well human and machine perform when translating distinct genres and quantify the differences clearly, the translation will be evaluated and scored based on Yans principle. 2. Participants Ten senior students majoring in English are randomly chosen as participants of the experiment. All of them have attended translation courses in the previous semesters, which means their English proficiency is high enough to check out the basic errors in vocabulary and grammar and distinguish the quality of the translation. Participants could accomplish the evaluation task independently at any place within the same period of time. It is suggested that participants should not read the materials twice and change the score they have already given. In this case the result will be more reliable. 3. Materials Due to the fact that participants are all Chinese, it is natural that they are in Chinese thinking patterns and the original materials in Chinese will not lead them to misinterpretation of the sentence meaning, so the research starts with C-E translation instead of E-C translation. The materials are selected from a bilingual website named www.EN84.com, which is specialized in providing superior English learning resources. The research mainly focuses on the differences between human translation and machine translation in two distinct genres, political speech and literature. The reason for choosing the two types of passages is that they have their own features, the former official with similar patterns while the latter creative beyond imagination, but both contain some certain emotional feelings, thus reasonable in testing the ability of human and machine in processing them. Remarks by H.E. Wang Yi at the MFA 2020 New Year Reception and Liang Shiqius prose Snow are suitable for the study. Since the work of making a judgement after reading long pieces of articles is time-consuming and probably adds errors to the result, the materials are excerpted and divided into 10 small paragraphs, with 5 for each genre. As for the machine translation tools, based on the research conducted by previous experts (Tian, 2016), statistical translation systems Google Translate and Baidu Translate are better choices than others because they are simple to operate and their accuracy is relatively high. The excerpted paragraphs will be input into the two translation systems and then the translations are generated. To design a logical evaluation paper, three points should be paid attention to. First of all, the human and machine translations of each paragraph are listed on one paper without showing participants which one is translated by human or machine to make sure stereotypes do not interfere with the result. Besides, the sequence of the two translations for each paragraph is randomly arranged so that evaluators can avoid getting trapped in the mindset that the human or machine version always goes first. The last one is that there is a clear differentiation in difficulty when selecting sentences. 4. Procedures The evaluation paper is sent to participants online and a maximum of thirty minutes is allowed to finish this task. Ten participants will be required to score each translation from three aspects and the grade ranges from zero to five (see Appendix for the detailed evaluation paper). Then, after collecting the data, the researcher will calculate the average point for each type of articles in terms of each standard and compare them to draw a conclusion.