Yuelu Mountain Universities Science City 2022

国家社会科学基金项目课题论证活页 Leaflet of project demonstration of National Social Science Foundation 课题名称：《鲁迅全集》基于**软件的机器翻译译后编辑与研究 Project title: Post-translation editing and research of The Complete Works of Lu Xun based on ** software 商务伙伴 1：网易有道 Business Partner 1: NetEase Youdao 商务伙伴 2：...... Business Partner 2:......

0. 题解 0. Answer key

1.选题依据（国内外相关研究的学术史梳理及研究动态；本课题相对于已有研究的独到学术价值和应用价值等） 2.Basis of selected topic (academic history and research trends of relevant research at home and abroad; The unique academic value and application value of this subject compared with the existing research) 机器翻译（Machine Translation）是使用计算机将一种语言符号转换成另一种语言符号，自20世纪30年代问世开始，经历了基于规则（rule-based）、基于实例（example-based）、基于统计（statistic-based）和基于不同方法应用（multiple methods application-based）的四个发展阶段，基于规则和基于统计的混合式机译系统是如今机器翻译主流模式之一，能很好地规避单一模式可能引发的弊病。随着AI技术的不断发展，机器翻译也在不断完善和发展，能够帮助人们快速、准确地搜索并记忆专业术语（相对于人工翻译而言此项工作既耗时耗力，反复比对后结果还不尽如人意）；机器翻译耗时少、处理速度快、工作环境要求低（只需电脑和翻译软件即可，而人工翻译需要一个安静无干扰的工作环境，译者的身体和精神状态无疑也会影响翻译的质量）；能够帮助减少甚至消除人工翻译的体力和脑力负荷也能大大降低翻译的成本；机器翻译都是多语“人才”，只要人们不断丰富平行语料库文本的语种，所有的语言互译都是可以实现的，而从事翻译的人们能熟练掌握3 - 4种语言互换就已经是凤毛菱角了，更不用说所有的语言互换了。在认识机器翻译强大优势的同时，我们也不能忽视现有机器翻译，因为自身的不足呈现出一些通用的弊病：各种误译、错译、漏译、语序错误、句法结构误译、句法不同等晦涩难懂的译文；机器翻译对于词语的处理仍停留在字面和浅层意义对应，无法实现词语深层意义和语用意义对应；机器翻译只能解读程式化和信息性文本的语义内容，还无法解读表情性文本中蕴含的情感和态度等等。 Machine Translation is the use of computers to convert symbols of one language into symbols of another language. Since its inception in 1930s, It has experienced four development stages of rule-based, example-based, statistic-based and multiple methods application-based. Rule-based and statistics-based hybrid machine translation system is one of the mainstream models of machine translation, which can well avoid the disadvantages caused by a single model. With the continuous development of AI technology, machine translation is also improving and developing, which can help people search and remember professional terms quickly and accurately (compared with human translation, this work is time-consuming and time-consuming, and the results after repeated comparison are not satisfactory). Machine translation work less time consuming, fast processing speed, low environmental requirements (by computer and translation software, and human translation requires a quiet working environment without interference, and the physical and mental state of the translator will definitely affect the quality of the translation), can help to reduce or even eliminate human translation of physical and mental load can also greatly reduce the cost of translation; Machine translation is a multilingual "talent", as long as people continue to enrich the language of the parallel corpus text, all language translation can be realized, and the translator can master 3-4 language exchange is already a chicken's head, let alone all the language exchange. While recognizing the powerful advantages of machine translation, we should not ignore the existing machine translation, because its own shortcomings present some common shortcomings: various mistranslations, mistranslations, omissions, word order errors, syntactic structure mistranslations, syntax differences and other obscure translation; Machine translation still deals with the correspondence between literal and superficial meanings, and fails to realize the correspondence between deep meanings and pragmatic meanings. Machine translation can only interpret the semantic content of stylized and informative texts, but not the emotions and attitudes contained in expressive texts. 针对机器翻译现有的问题，人们开发了各种机器翻译译前、译后的人工编辑软件和平台，诸如谷歌翻译工具包、Trados Studio、Systran、MateCat和Memsource，通过人工参与编辑解决机器翻译中出现的各种错误。但通过对比机器翻译译后编辑和人工翻译的速度，人们发现当前译后编辑在效率方面优势不明显，二者的错误类型相似（误译、省略、语法错误等）、译后编辑存在较多过度编辑的情况（语序调整、人称代词删除等）等问题。 In view of the existing problems of machine translation, people have developed a variety of human editing software and platforms before and after machine translation, such as Google Translation Toolkit, Trados Studio, Systran, MateCat and Memsource, which solve various errors in machine translation through human participation. However, by comparing the speed of machine translation and human translation, it is found that the current post-translation editing has no obvious advantage in efficiency, and the two types of errors are similar (mistranlation, omission, grammatical errors, etc.), and the post-translation editing has a high degree of over-editing (word order adjustment, personal pronoun deletion, etc.). 2016年，国务院发布的《“十三五”国家科技创新规划》明确指出重点开发人工智能等技术；2017年，“人工智能”首次出现在《政府工作报告》中；同年发布的《新一代人工智能发展规划》提出，“人工智能”将“无时不有、无处不在”，并将其列为一项重要的国家战略。国家政策大力支持发展人工智能，广大社会群体对基于人工智能技术的机器翻译需求日益增加，也对机器翻译提出了更高的要求，同时也带来了无限的发展机遇。俞凯（2019）指出要提升我国人工智能国际竞争力的迫切需求，新一代人工智能关键共性技术的研发部署要以算法为核心。 In 2016, the 13th Five-Year National Science and Technology Innovation Plan issued by The State Council clearly pointed out that the development of artificial intelligence and other technologies; In 2017, "artificial intelligence" was first mentioned in the Government Work Report. The Development Plan for the Next Generation of ARTIFICIAL Intelligence, released in the same year, proposed that "ARTIFICIAL intelligence" would be "everywhere, all the time" and identified it as an important national strategy. The national policy strongly supports the development of artificial intelligence, and the general social groups have an increasing demand for machine translation based on artificial intelligence technology, which also puts forward higher requirements for machine translation, and also brings infinite development opportunities. Yu Kai (2019) pointed out that in order to improve the international competitiveness of China's ARTIFICIAL intelligence, the r&d and deployment of the key generic technologies of the new generation of artificial intelligence should take algorithms as the core. 本研究旨在通过**软件对《鲁迅全集》（共计700万字）实施中德、中英机器翻译，再进行译后编辑，总结机器翻译常见错误类型以及译后编辑处理时常见的问题，并系统总结机器翻译产生错误的原因、采用的策略和方法，继而探讨采取什么策略和方法可以提高机器翻译的准确度？使用什么方法能够减少译后编辑的人工参与度？如何构建适用于任意语对转换的翻译模型？此外，还可以协助工程师完善以自然语言理解为核心的认知计算理论和方法（如算法的创新与组合），提高计算力来推进深度学习（文本学习与深度学习结合），在国内推动中华文化传播与中华典籍外译工作，提升我国人工智能的国际竞争力，在国际上加速不同语际间的文化传播，实现新时期机器翻译发展的大跃进。在此过程中，将通过为期2年左右的时间，创建高质量的中德、中英双语平行语料库（《鲁迅全集》为文本，语料库总字数约1400万字），采用语料库语言学法、跨学科研究法和数量研究法等研究方法。 This study aims to implement machine translation from Chinese to German and Chinese to English for the Complete Works of Lu Xun (7 million words in total) by ** software, and then conduct post-translation editing, summarize the common types of errors in machine translation and the common problems in post-translation editing, and systematically summarize the causes of errors in machine translation, strategies and methods adopted. Then, what strategies and methods can be adopted to improve the accuracy of machine translation? What can be done to reduce human involvement in post-translation editing? How to construct a translation model for any Italian pair conversion? In addition, can also help engineers improve the natural language understanding is the core of cognitive computational theory and methods, such as innovation and composition) of the algorithm, improve the computing power to promote deep learning (text combine learning and deep learning), outside the domestic promoting the dissemination of Chinese culture and the Chinese classics translation work, enhance the international competitiveness of China's artificial intelligence, It will accelerate the cultural communication between different languages internationally and achieve a great leap forward in the development of machine translation in the new era. In this process, a high-quality Bilingual parallel corpus of German and Chinese and English will be created in about two years (the Complete Works of Lu Xun is the text, with a total of about 14 million words in the corpus), and research methods such as corpus linguistic method, interdisciplinary research method and quantitative research method will be adopted.

1.1国内外学术动态 1.1 Academic trends at home and abroad 1.1.1机器翻译国内外学术动态 1.1.1 Domestic and foreign academic trends of MACHINE translation 机器翻译在翻译界是一门新兴学科，也是一项新兴技术。依据ISO/DIS 17100: 2013标准的定义，机器翻译是“使用计算机系统将文本或语音从一种自然语言自动翻译为另一种语言”（MT, automated translation of text or speech from one natural language to another using a computer system）。20世纪30年代，法国科学家G・B・Artsouni最早提出机器翻译的设想；1946年，英美两位工程师A・D・Booth和Warren・Weaver在讨论计算机可能的应用范围之时，提出利用计算机进行语言翻译；1949年，Weaver首次提出使用计算机进行翻译的思想，提出了避免出现“字对字”翻译的四条具体原则。Bar Hillel于1952年组织召开了第一次机器翻译大会，机器翻译是一门跨学科的研究，各国纷纷加入机器翻译研究的队列。 Machine translation is a new subject and technology in the field of translation. According to the ISO/DIS 17100: Definition of 2013 Standard, Machine translation is "the automatic translation of text or speech from one natural language into another using a computer system" (MT, Automated translation of text or speech from one natural language to another using a computer system). In the 1930s, the French scientist G. B. Artsouni first proposed the idea of machine translation; In 1946, two British and American engineers A. D. Booth and Warren Weaver proposed the use of computer for language translation when discussing the possible application range of computer. In 1949, Weaver first put forward the idea of using computers for translation, and put forward four specific principles to avoid "word-for-word" translation. Bar Hillel organized and held the first Machine translation Conference in 1952. Machine translation is an interdisciplinary study, and countries have joined the queue of machine translation research. 1954年，美国乔治敦大学和IBM公司开展了世界第一次计算机翻译；1950至1960年间，机器翻译的研究主要针对理论语言学中的句法分析，建立了多种基于句法分析的机器翻译模型，然而研究发现基于句法分析的机器翻译并不能生成高质量的译文（Bar Hillel, 1958; 1959）。为了进一步推动机器翻译的发展，人们建立了自动语言处理咨询委员会（ALPAC，The Automatic Language Processing Advisory Committee）。而该机构成员较少关注机器翻译的学术价值以及研究发展的潜力，且仅关注英俄双语互译，得出的译文质量的结论不全面，没有对报告进行更深刻的研究，因此ALPAC 出示的报告信用力度与有效度存在不足。（Hinton et al, 2012） In 1954, Georgetown University and IBM launched the world's first computer translation; From 1950 to 1960, researches on MACHINE translation mainly focused on syntactic analysis in theoretical linguistics, and various models of machine translation based on syntactic analysis were established. However, researches found that syntactic machine translation based on syntactic analysis could not produce high-quality translation (Bar Hillel, 1958; 1959). In order to further promote The development of machine translation, people set up The Automatic Language Processing Advisory Committee (ALPAC). However, the members of this organization paid little attention to the academic value and research development potential of machine translation, and only paid attention to the bilingual translation between English and Russian. The conclusion of translation quality was not comprehensive, and they did not conduct a deeper study on the report. Therefore, the credibility and validity of the report presented by ALPAC were insufficient. (Hinton et al., 2012) 1966年，自动语言处理咨询委员会发布题为《语言与机器》的报告，给机器翻译带来了很大的影响：一是导致机器翻译的应用投资以及技术研究大范围停滞，但在加拿大、法国、欧盟等国家和地区的机器翻译研究却表现突出。加拿大于1965年在蒙特利尔成立了机器翻译研究中心，建立了TAUM机器翻译系统，主要从事英法双语机器翻译研究与实践（Poibeau, 2017）；20世纪70年代中期，Vauquois将之前应用的翻译系统进行了一定程度的改造。由于各个国家的语言翻译需求不一，欧盟与在美国诞生的第一个商业机器翻译系统开发商Systran合作，开展欧盟成员国语言的机器自动翻译。 In 1966, the Automatic Language Processing Advisory Committee issued a report entitled "Language and Machine", which brought great influence to machine translation. First, the application investment and technical research of machine translation largely stopped, but the research of machine translation in Canada, France, the European Union and other countries and regions showed outstanding performance. Canada established the Machine Translation Research Center in Montreal in 1965 and established TAUM Machine Translation System, which is mainly engaged in the research and practice of English-French bilingual machine translation (Poibeau, 2017). In the mid-1970s, Vauquois partially adapted the translation system previously used. The European Union has teamed up with Systran, the first commercial machine translation system developed in the United States, to automate machine translation of eu member states' languages, as countries have different needs for language translation. 1990年国际计算语言学大会在芬兰召开，辛顿认为这次会议开启了基于大规模平行语料库的统计机器翻译时代（Hinton et al，2012）。此前，机器翻译经历了基于词典和手工规则的机器翻译系统（rule-based translation），基于实例的机器翻译系统（example-based translation），发展到基于统计的机器翻译系统（statistical machine translation）。 In 1990, the International Congress of Computational Linguistics was held in Finland. Hinton believed that this conference opened the era of statistical machine translation based on large-scale parallel corpora (Hinton et al., 2012). Previously, MACHINE translation has experienced rule-based translation system based on dictionaries and manual rules, example-based translation system, To develop statistical machine Translation system based on statistics. 随着深度学习的不断发展，神经网络机器翻译逐渐成为翻译领域的主力军，成为翻译领域最强大的算法。这种最先进的算法是深度学习的一项应用，其中大量翻译完成的句子的数据集被用于训练能够在任意语言对之间的翻译模型。 With the continuous development of deep learning, neural network machine translation has gradually become the main force and the most powerful algorithm in the field of translation. This state-of-the-art algorithm is an application of deep learning, in which a large data set of translated sentences is used to train a translation model capable of between arbitrary language pairs. 而国内机器翻译的发展较晚，机器翻译研究则是从1956年开始的；1959年，中科院语言所和计算所共同研制了俄-汉翻译系统；20世纪70年代，我国开始重视机器翻译；20世纪80年代中期到90年代初期，是我国机器翻译发展的三个重要时期，分别于1987年和1992年，发明了KY-1英汉翻译系统以及IMT/EC863英汉机译系统；从90年代初到现在，我国机器翻译研究蓬勃发展。 However, the development of machine translation in China is relatively late, and the research on machine translation started in 1956. In 1959, the Russian-Chinese translation system was jointly developed by the Institute of Language and Computer Science of the Chinese Academy of Sciences. In the 1970s, China began to attach importance to machine translation. From the mid-1980s to the early 1990s, there were three important periods in the development of Machine translation in China. In 1987 and 1992, KY-1 english-Chinese translation system and IMT/EC863 English-Chinese machine translation system were invented respectively. From the early 1990s to the present, machine translation research in China has developed vigorously. 詹卫东（2002）讨论了服务于机器翻译的语言研究需要关注的问题，需要重视研究手段和研究方式的创新，结合新科技开发语言研究平台。陈丹（2013）分析了语料库的概念和某些种类语料库在翻译活动中具体的实用情况，指明语料库对于翻译研究的意义所在。屈亚媛、周玉梅（2016）以《黄帝内经·素问》养生篇章中双字格术语为文本，对比中医专家李兆国和SYSTRAN的译文，分析机译翻译错误常见的形式及成因，旨在通过人工改进机读养生术语词典来优化中医养生术语机译数据库。胡开宝、李毅（2016）指出机器翻译的主要特征包括五个方面：自动化、机械性、以语句为翻译单位、二度摹仿和语境制约。机器翻译和人工翻译是相辅相成、相互促进的关系，并非矛盾和零和的关系，还提到机器翻译最终不会完全取代人工翻译。丁亮（2017）针对当前机器翻译的领域自适应方法缺乏较明确的领域标签，仅仅将数据粗略划分为领域内(in-domain)和领域外(out-domain)两类，提出利用中途分类号作为领域标签，使用论文关键词和科技词系统等知识组织构建领域知识库对汉语句子进行自动领域标注以及训练卷积神经网络的深度学习的领域标注方法，生成效果更佳的领域标注器，并通过机器翻译的测试集获取领域标签集合筛选其训练数据。冯志伟（2018）指出人工智能与机器翻译关系密切，分析了基于规则、统计和神经的机器翻译原理和方法，强调机器翻译和人工智能目前还不够成熟，仍处于发展初期。吴戈（2019）提出机器翻译是一种“语言游戏”，引发关于机器翻译的语义和数据之争，总结哲学和科学的方法并总是矛盾的，只有正确理解二者之间的关系才能帮助解决语义方面的问题。陈伟（2020）指出从技术逻辑层面来看，机器翻译取代人工翻译呈现必然的趋势；机器翻译与人工翻译的根本差异在于主体不同，其本质为翻译主体性在场或缺失；翻译本性的体现包括意义的开放性、表达的创造性和翻译的政治性，而这三点正是未来人工翻译的着力点。王湘玲（2021）采用眼动追踪、键盘记录、反省报告和问卷调查多元互证的方法对比人工翻译和译后编辑在处理隐喻表达时承受的认知负荷和译文质量，发现采用译后编辑模式可以大大减少译者认知负荷并提高译文质量。 Zhan Weidong (2002) discussed the issues that need to be paid attention to in language research for machine translation, and the need to attach importance to the innovation of research methods and methods, and develop language research platforms based on new technologies. Chen Dan (2013) analyzed the concept of corpus and the specific practical situations of some kinds of corpus in translation activities, and pointed out the significance of corpus for translation studies. Qu Yayuan and Zhou Yumei (2016) analyzed the common forms and causes of translation errors in machine translation by comparing the translations of Li Zhaoguo and SYSTRAN, traditional Chinese medicine experts, with the text of double-character terms in the health preservation chapter of Yellow Emperor's Inner Classic · Suwen, aiming to optimize the machine translation database of TCM health preservation terms by manually improving the machine reading dictionary of health preservation terms. Hu Kaibao and Li Yi (2016) point out that the main features of machine translation include five aspects: automation, mechanicity, taking sentence as translation unit, second-degree imitation and context restriction. Machine translation and human translation complement and promote each other rather than contradict and zero-sum. It is also mentioned that machine translation will not completely replace human translation in the end. According to Ding Liang (2017), the current domain adaptive methods of MACHINE translation lack clear domain labels and only roughly divide data into in-domain and out-of-domain categories, and propose to use the midway classification number as domain labels. Use key words and word of science and technology system, and other areas of the knowledge organization to build knowledge base of Chinese sentences automatically with the depth of the convolutional neural network learning and training in the field of field of labeling method, generate better field annotator, and through the test of the machine translation label collection screening of the training data set to obtain field. Feng Zhiwei (2018) pointed out the close relationship between artificial intelligence and machine translation, analyzed the principles and methods of machine translation based on rules, statistics and nerves, and emphasized that machine translation and artificial intelligence are still immature and in the early stage of development. Wu Ge (2019) proposed that MACHINE translation is a kind of "language game", which causes semantic and data disputes about machine translation. The methods of summarizing philosophy and science are not always contradictory. Only a correct understanding of the relationship between the two can help solve semantic problems. Chen Wei (2020) points out that from the perspective of technical logic, machine translation will inevitably replace human translation. The fundamental difference between machine translation and human translation lies in the presence or absence of translation subjectivity. The nature of translation includes openness of meaning, creativity of expression and political nature of translation, which are the focus of future artificial translation. Wang Xiangling (2021) compared the cognitive load and translation quality of human translators and post-editors when processing metaphorical expressions by using multiple cross-verification methods such as eye movement tracking, keylogging, introspective report and questionnaire survey, and found that post-editing mode can greatly reduce translators' cognitive load and improve translation quality.

1.1.2译后编辑国内外学术动态 1.1.2 Post-editing domestic and foreign academic developments 了解译后编辑（PE, post-editing），译后编辑则是“检查和修正机器翻译的输出（to check and correct MT output）”（ISO, 2014）。Allan（2003）指出从1947年英国工程师A. D. Booth和美国科学家W. Weaver首次进行俄英机器翻译到现代翻译技术高速发展的今天，机器翻译的质量大大提高，但鉴于语言结构、修辞手法、逻辑思维、文化差异、一词多义等现象，我们仍需对机器翻译的译文进行修改和加工，即译后编辑来提高机器翻译质量，如修改语言错误、提高表达准确性与译文可读性等。 Understanding PE, post-editing is "to check and correct MT output" (ISO, 2014). Allan (2003) pointed out that from 1947, British engineer A. D. Booth and W. Weaver maiden, Russia, Britain, machine translation to modern technology rapid development today, greatly improving the quality of machine translation, but given the structure of language, rhetoric and logic thinking, cultural differences, the phenomenon such as polysemy, we still need to modify for the machine translation and processing, namely the editor to improve the quality of machine translation, translation Such as correcting language errors, improving expression accuracy and translation readability. 20世纪30年代初法国工程师G. B. Artsouni最早提出机器翻译的设想，1947年英国工程师A. D. Booth和美国科学家W. Weaver首次提出利用计算机进行翻译。W. Weaver于1949年发表《翻译》备忘录，并提出机器翻译的思想。1954年，美国乔治敦大学和IBM公司进行世界第一次计算机俄英翻译实验，随后前苏联及欧洲国家因为军事或经济发展需要，非常重视机器翻译研究，机器翻译呈现热潮。同年，Victor H. Yngve（1954）简要介绍译后编辑的概念、任务和作用，指出机器翻译与人工合作将成为未来机器翻译行业发展的特点。Miller和Beebe-Center（1956）介绍五种机器翻译质量评价方法，即主观评判、单词评判（单词数及顺序）、信息传译评判、阅读理解评判和语义评判，并分析各评价方法的优劣，同时指出并非所有文学作品都可以通过机器成功翻译。1966年由美国政府资助的语言自动处理咨询委员会（Automatic Language Processing Advisory Committee, ALPAC）通过调查研究发布了报告——《语言与机器：翻译和语言学中的计算机》（Language and Machines: Computers in Translation and Linguistics），指出机器翻译质量粗糙，未达到预期效果，建议专注字典开发以及译后编辑，使机器翻译资助锐减。（ALPAC, 1966） In the early 1930s, French engineer G. B. Artsouni first proposed the idea of machine translation. In 1947, British engineer A. D. Booth and W. Weaver first proposed the use of computers for translation. W. Weaver published the memorandum "Translation" in 1949 and put forward the idea of machine translation. In 1954, Georgetown University and IBM conducted the world's first computer Russian-English translation experiment. Later, the former Soviet Union and European countries attached great importance to machine translation research due to the needs of military or economic development, and machine translation became a boom. In the same year, Victor H. Yngve (1954) briefly introduced the concept, task and role of post-translation editing, and pointed out that machine translation and human cooperation would become the characteristics of the future development of machine translation industry. Miller and Beebe-Center (1956) introduced five quality evaluation methods of machine translation, namely subjective evaluation, word evaluation (word number and order), information translation evaluation, reading comprehension evaluation and semantic evaluation, analyzed the advantages and disadvantages of each evaluation method, and pointed out that not all literary works can be successfully translated by machine. In 1966, the Automatic Language Processing Advisory Committee (ALPAC), funded by the U.S. government, published a report "Language and Machines: Language and Machines: translation and Linguistics Computers in Translation and Linguistics, pointed out that the quality of machine Translation is rough and does not achieve the expected effect, and suggested that focus on dictionary development and post-translation editing, so that the funding for machine Translation decreased sharply. (ALPAC, 1966) Orr和Small（1967）对比人工翻译、机器翻译和机器翻译译后编辑模式下俄英翻译速度、效率和准确度方面的差异，发现人工翻译效果优于译后编辑，在三个方面都超越了机器翻译，速度与效率的差异尤甚，而机器翻译与译后编辑的差异相较于人工翻译与译后编辑的差异更大。鉴于译后编辑压力大、费用高，二者仍建议发展机器翻译。Van Slype（1978）在欧盟引进Systran机器翻译后，写了一份报告分析机器翻译、人工翻译和译后编辑译文的可读性，分别为78%，98%和98%，大多数机器翻译用户认为某些情况下的机器翻译译文质量是可以被接受的。来自加拿大通用汽车有限公司的Sereda（1982）指出译后编辑是机器翻译系统中一个非常重要的因素，译后编辑的成功与否直接关系到机器翻译产生的经济价值大小，还谈到影响译后编辑功能的因素包括机器翻译系统的语言能力、源语质量、专业术语存储、个人及机器翻译能力等。Green（1982）将机器翻译错误分成简单的小型错误、耗时的大型错误以及“灰色”错误，并建议译后编辑在面对之前出现过的文本时，可以提前编辑那些有代表性的数据以避免犯相同的错误，从而降低译者的烦恼指数。Laurian（1984）逐渐侧重快译后编辑，将其按照错误修正优先等级分成“必须修改错误”、“可能需要修改错误”和“不该出现的错误”三类，并指出各自的区别，提出译后编辑既不是修改，也不是更正，更不是改写。作者一共收集了11种错误类型，帮助那些没有经过太多语言培训的人群判断文本是否适合译后编辑，这就是后来的机器翻译可行性研究。Elizabeth Wagner（1985）从1981年开始收集数据了解当时使用快译后编辑服务的人群和目的，并分析翻译软件Systran的弊端在于译“词”而非译“意”。Vasconcellos（1985）认为理想的译后编辑应该是一个职业译员，而非某个主题专家，但她也指出译者对于机器翻译的态度是决定译文质量的最关键因素。Schneider（1989）针对西门子公司对译后编辑的培训谈到译后编辑起始阶段翻译效率会降低，经过几个月到一年不等的时间，使用者反馈效率大大提高，语言转换时间逐渐减少。 Orr and Small (1967) compared the differences in speed, efficiency and accuracy of Russian-English translation in the mode of manual translation, machine translation and machine translation post-editing, and found that the effect of manual translation was better than that of post-editing, and exceeded that of machine translation in all three aspects, especially in speed and efficiency. The differences between machine translation and post-translation editing are greater than those between human translation and post-translation editing. In view of the high pressure and cost of post-translation editing, they still recommend the development of machine translation. Van Slype (1978), after the introduction of Systran machine translation in the EU, wrote a report analyzing the readability of machine translation, human translation and post-edited translation, with results of 78%, 98% and 98% respectively. Most machine translation users believe that the quality of machine translation in some cases is acceptable. Sereda (1982) from General Motors Of Canada pointed out that post-editing is a very important factor in machine translation system, and the success of post-editing is directly related to the economic value generated by machine translation. The factors that affect the post-translation editing function include the language ability of the machine translation system, the quality of the source language, the storage of professional terms, and the ability of individual and machine translation. Green (1982) divided machine translation errors into simple small errors, time-consuming large errors and "gray" errors, and suggested that post-translation editors should edit representative data in advance to avoid making the same mistakes when faced with previously appeared texts, so as to reduce the translator's annoyance index. Laurian (1984) gradually focused on post-fast translation editing and divided it into three categories according to the error correction priority: "errors that must be modified", "errors that may need to be modified" and "errors that should not occur", and pointed out their differences. Post-translation editing was neither modification, correction nor rewriting. The author collected a total of 11 error types to help those without much language training judge whether a text is suitable for post-translation editing, which became the feasibility study of machine translation. Elizabeth Wagner (1985) began to collect data in 1981 to understand the people and purposes of quick translation and post-editing services at that time, and analyzed that the drawback of translation software Systran lies in the translation of "words" rather than "meanings". Vasconcellos (1985) believed that the ideal post-translation editor should be a professional translator rather than a subject matter expert, but she also pointed out that the translator's attitude towards machine translation is the most critical factor to determine the quality of translation. Schneider (1989) pointed out that the translation efficiency of post-editing would decrease in the initial stage of post-editing. After several months to a year, users' feedback efficiency would be greatly improved, and the time of language conversion would gradually decrease. Dorothy Senez（1998）分析译后编辑与翻译的区别，指出前者在增加译文数量、有效利用现有工具和降低翻译成本方面前景广阔。Krings, Hans P.（2001）采用实证研究方法完成博士后研究报告，说明译后编辑即文本修改的过程。同年，该报告被Geoffrey S. Koby从德文译成英文并出版，成为译后编辑领域中最早的一本专著。Allen（2003）讨论机器翻译译后编辑本地化进程、译后编辑原则以及新开发的译后自动编辑模型。Guerberof（2009）通过实验研究机器翻译片段与翻译记忆系统中模糊匹配片段在效率和质量上的关系，发现使用机器翻译的译者工作更高效、译文质量更高，也指出具有翻译技术经验的译者更高效，但译文质量不受影响。Midori Tatsumi（2010）采用定性和定量研究分析发现句子结构、文本类型、术语使用、译后编辑模式与译后编辑行为等因素共同影响译后编辑人员的编辑数量，该研究有助于业界更好地使用机器翻译系统，也能帮助译后编辑人员提高译后编辑技巧和策略。翻译自动化用户协会（TAUS, Translation Automation User Society）（TAUS, 2010）指出译后编辑指导原则的内容涵盖未做译后编辑的原始机器译文和人们对翻译内容的最终预期质量，个人可以根据顾客需要和机器翻译文本质量来选择相应的指导原则。Garcia（2011）指出当下翻译记忆工具能帮助译员将数据库不能匹配的词条存入记忆库，通过实验对比译后编辑与人工翻译发现译后编辑在效率方面优势较小，在翻译质量方面优势非常明显，并从双语方向、文本难度和译者水平层面分析其对译文质量的影响，讨论译者是否应该考虑使用译后编辑替代传统翻译模式。 Dorothy Senez (1998) analyzed the difference between post-translation editing and translation and pointed out that the former has a broad prospect in increasing the number of translations, effectively utilizing existing tools and reducing translation costs. Krings, Hans P. (2001) used empirical research methods to complete the post-doctoral research report, explaining that post-translation editing is the process of text modification. In the same year, the report was translated from German into English by Geoffrey S. Koby and published, making it the first monograph in the field of post-translation editing. Allen (2003) discussed the localization process of post-translation editing in machine translation, post-translation editing principles and the newly developed automatic post-translation editing model. Guerberof (2009) conducted an experimental study on the relationship between the efficiency and quality of machine translated fragments and fuzzy matched fragments in translation memory system, and found that translators using machine translation work more efficiently and translation quality is higher. It also pointed out that translators with translation technology experience are more efficient, but the quality of translation is not affected. Midori Tatsumi (2010) used qualitative and quantitative research to analyze and find that sentence structure, text type, use of terms, post-editing mode and post-editing behavior and other factors jointly affect the number of post-editing staff. This research is conducive to better use of machine translation system in the industry. It can also help post-editors to improve their post-editing skills and strategies. Translation Automation User Society (TAUS, TAUS, 2010) points out that the post-editing guidelines cover the original machine Translation without post-editing and the final expected quality of the translated content. Individuals can choose appropriate guidelines based on customer needs and machine-translated text quality. Garcia (2011) pointed out that the present translation memory tools can help the interpreter in memory, the database can't match the entry through experiment contrast translation editing and artificial translation found after editing in efficiency advantage is small, very obvious advantages in terms of quality of translation, and from the bilingual direction, difficulty of text and translator level level to analyze its impact on the quality of the translation, Discuss whether translators should consider using post-editing to replace traditional translation models. 2013年第十四届机器翻译峰会（Machine Translation Summit XIV）在法国召开，期间举行了“第二届译后编辑技术与实践研讨会”（The 2nd Workshop on Post-editing Technologies and Practice）重点探讨译后编辑界面和功能设置、译后编辑效率、译后编辑评测工具、在线译后编辑环境、机器翻译与译后编辑技术推广、单双语译后编辑效果对比与编辑能力差异、译前编辑对译后编辑的推动作用、人工翻译与机器翻译译后编辑模式的译文质量对比、译后编辑处理等议题。（Sharon O’Brien, 2013） 2014年颁布并实施了关于机器翻译译后编辑的草案标准ISO18587，指出译后编辑三个阶段具体要求，即准备阶段、译后编辑阶段和译后编辑后处理阶段，描述译后编辑人员需具备的能力和资质，将译后编辑输出分为快译后编辑（Light post-editing）和完全译后编辑（Full post-editing）两个级别，指出对应特点，该标准得到各行业的广泛认可。ISO 18587（2017）区分了完全译后编辑和快译后编辑的差异，探讨快译后编辑方式是否也能满足顾客需求，并指出必须对译后编辑从业人员的机器翻译译后编辑能力进行培训。Koponen（2017）首次对比分析英芬语机器翻译后必须修改的地方以及译后编辑人员自己所做的修改，发现大部分编辑都是正确的，但也有34%的编辑是多余的。研究表明这两种语言译后编辑采取的调整词序、删除人称代词是多余的，这对于将来译后编辑实践和培训颇有指导意义。Lommel（2018）指出单靠语言学家无法满足当代人们对快速和低价翻译的需求，拥抱机器翻译的语言服务商已经取得了飞速的发展，介绍机器增强翻译（神经机器翻译）的定义和组成内容，详细探讨适应性机器翻译和内容充实自动化、该领域的技术供应商以及机器增强翻译的未来。Sakamoto（2019）用定性和定量研究方法引用布迪厄文化资本理论“资本”、“场域”和“惯习”的概念，分析译者为何不愿从事译后编辑的原因，该研究为机器翻译股东和翻译工作教育者理解译员与译后编辑所处的社会机构提供了概念化的工具。Seung（2020）旨在探寻指导机器翻译译后编辑的方法论思想以改进译文风格，尤其是针对句法和语义上有巨大差异的英韩等语对。第一步整理出双语写作中语言和文化的不同之处，第二步了解译员对于写作风格的不同命名，第三步将人工翻译中使用的策略引入到机器翻译译后编辑，研究如何将这些策略植入到英韩机器翻译译后编辑中，从而改善译文风格。该研究为改善译后编辑风格指导标准奠定基础，为改善译文风格积累人工译后编辑数据，从而帮助打造能满足不同顾客需求的译后编辑自动化系统。Koponen（2021）指出人脑修改译文与机器修改译文之间的界限正在逐渐消失，政府或公司翻译部门、翻译公司、文学出版部门、志愿者部门以及人工和机器修改培训领域研究，具体包括基于调查、访谈、击键记录的实证研究，更侧重从理论层面探寻翻译与译后编辑的传统区别。错误修订和译后编辑研究涉及8种语言，研究话题涉及译者和修订者关系、非专业译员的修订行为与译后编辑。 The 14th Machine Translation Summit XIV was held in France in 2013, The 2nd Workshop on post-Editing Technologies and Editing was held Practice), the paper mainly discusses the translation after editing interface and feature set, the translation editor efficiency, translation evaluation tools, online editing environment after translation, machine translation and post-translation editing technology popularization, single and double language translation editing effect after contrast and editing ability difference, former editor of the translation editor role after the translation and human translation and machine translation quality of edit mode after contrast , post-translation editing processing and other issues. (Sharon O 'Brien, 2013) In 2014, the draft standard ISO18587 on post-translation editing of machine translation was issued and implemented, which pointed out the specific requirements of three stages of post-translation editing, namely, the preparation stage, the post-translation editing stage and the post-translation editing post-processing stage, and described the abilities and qualifications of post-translation editing personnel. The output of post-editing is divided into two levels, Light post-editing and Full post-editing, pointing out the corresponding characteristics, the standard has been widely recognized by various industries. ISO 18587 (2017) distinguishes between full post-translation editing and fast post-translation editing, discusses whether fast post-translation editing can also meet customer needs, and points out that post-translation editing practitioners must be trained in machine translation post-translation editing ability. Koponen (2017) made a comparative analysis for the first time on the necessary modifications in English-Finnish machine translation and the modifications made by the editors themselves, and found that most of the edits were correct, but 34% of the edits were redundant. The research shows that the adjustment of word order and the deletion of personal pronouns are unnecessary, which is of great significance to the practice and training of post-translation editing in the future. Lommel (2018) points out that linguists alone cannot meet the needs of contemporary people for fast and cheap translation. Language service providers embracing machine translation have achieved rapid development. The definition and components of machine enhanced translation (NMT) are introduced. Explore in detail adaptive machine translation and content enrichment automation, technology providers in the field, and the future of machine enhanced translation. Sakamoto (2019) using qualitative and quantitative research methods refer to bourdieu cultural capital theory "capital" and "field" and "habitus" concept, analyze why the translator's reluctance to engage in translation after editing, the research of machine translation after shareholders and translation work educators understand the interpreter and edit social organization provides a conceptual tools. Seung (2020) aims to explore the methodological ideas guiding post-translation editing of machine translation in order to improve the translation style, especially for English and Korean language pairs with huge differences in syntax and semantics. The first step is to sort out the differences in language and culture in bilingual writing. The second step is to understand the different naming styles of writing by translators. The third step is to introduce the strategies used in human translation into machine translation and post-editing, and to study how to implant these strategies into English-Korean machine translation and post-editing so as to improve the translation style. This study lays the foundation for improving the guidance standards of post-translation editing style, and accumulates the data of post-translation editing for improving the style of translation, so as to help build an automatic post-translation editing system that can meet the needs of different customers. Koponen (2021) points out that the boundary between human brain modified translation and machine modified translation is gradually disappearing. Studies on government or company translation departments, translation companies, literary publishing departments, volunteer departments and human and machine modified training fields, specifically including empirical studies based on surveys, interviews and keystroke records. The traditional differences between translation and post-translation editing are explored from the theoretical level. The research on incorrect revision and post-editing involves eight languages, including the relationship between translator and reviser, non-professional translators' revision behavior and post-editing. 国内机器翻译研究于1956年开始，第二年，中国科学院语言研究所与计算机技术研究所合作开展俄汉机器翻译研究，译后编辑的研究主要围绕译后编辑工具的设计与开发。黄河燕（1994）提出新的智能译后编辑器的设计原理和算法以提高译后编辑效率。魏长虹（2007）指出机器翻译的各组成部分，译后编辑能帮助译员提高译文质量和人工校译效率，分析译后编辑的基本概念、译后编辑的必要性、译后编辑的手段、译后编辑需要修正的错误和译后编辑实施者五个方面的内容。2008年，格微公司设计“格微协同翻译平台”（GE-CCT），将机器翻译与译后编辑结合，并在商务翻译领域加以实践。李梅（2013）对汽车专业英汉翻译平行语料库英汉机器翻译中出现的规律性、典型性译文错误进行机器翻译二次加工，即通过译后编辑自动化过滤这部分错误来提高机器翻译速度和机器翻译质量。崔启亮（2014）探讨译后编辑的概念，分析译后编辑的应用与研究现状、推动译后编辑发展的动力、适合译后编辑研究的题材，并提出提高译后编辑质量和效率的行为准则。冯全功（2015）分别从译后编辑的行业需求、课程设置、编辑能力、教学及工具选择方面讨论译后编辑的培养方法，并指出在高校开设译后编辑课程能够增强学生职业竞争力，更好地满足语言服务业对译后编辑的需求。崔启亮（2015）针对英语科技文本译后编辑的案例分析机器翻译各种错误类型，包括过译、欠译、术语翻译、形式、格式、多译和漏译、冗余、词性判断、从句翻译、短语顺序、原文句子结构束缚，并总结译后编辑的特点。冯全功（2016）分析目前译后编辑研究的聚焦点，并指出将来研究发展趋势，如开发集成翻译环境、研发特定机器翻译系统、同译后编辑、应用人才培养以及产学研层面进行深度合作。冯全功（2018）尝试从认知、知识和技能三个维度构建译后编辑能力模型，论述每个维度的构成、表征和关联以及该模型对于译后编辑教学的启示作用。王湘玲（2018）通过对“翻译研究文献目录”与“机器翻译档案馆”两大数据库语料分析译后编辑过程及产品评估、影响效率因素、工具及人才培养的研究进程与研究方法，探索其研究趋势，为相关研究及翻译人才培养和学科建设提供新的研究视角和研究方法。王湘玲（2019）通过击键记录和问卷调查法，对比实验者在人工翻译和机器翻译译后编辑的翻译速度、译文质量、译者态度三方面的差异，从而指导未来机器翻译译后编辑人才培养。陆强（2019）研究依据翻译的准确性、完整性、术语一致等要求制定通用译后编辑质量规范指导机器翻译译后编辑项目，也指出可以根据具体项目特点调整项目质量规范，更好指导译后编辑工作。朱慧芬（2020）从纽马克翻译理论出发分析“八八战略”谷歌英译出现的错误，探索译者在文本、所指、衔接和自然层面的译后编辑原则，并指出隐含信息补充和“名从源主”的特色专有名词翻译原则。蔡强（2020）使用Google汉英机器翻译中国知网数据库中200篇科技论文摘要，发现机器翻译主要错误包括词汇、句法、逻辑和其他错误等，且错误频次依次递减，提出在词汇、句法、逻辑和标点使用方面进行译后编辑，丰富了科技文本机器翻译的译后编辑研究。赵涛（2021）指出人们还未能很好区分机器翻译和译后编辑的差异，有必要综合从多个角度去认识机器翻译的相对优势与固有缺点。机器翻译译后编辑可以根据文本类型和客户需求分类，适应不同的使用环境，还提到今后需要加大机器翻译译后编辑的教学研究工作，分别从知识、实践与技能层面培养和提高机器翻译译后编辑核心能力。李梅（2021）通过分析职业译员参加英汉机器翻译译后编辑测试结果研究原文对译后编辑时间与译文质量的关系，帮助人们理解人机协作、了解原文与译后编辑的关系以及提高译后编辑效力。陈胜（2021）通过对比分析汉语石油地质文献在7大主要国内外线上翻译平台英译，发现主要问题集中在词义、词性、词序、句子结构、断句、名词语法标记、搭配、标点符号、字母大小写、信息完整性等方面，提出采用“信达切”的原则开展译后编辑，寻找翻译处理速度与译文准确性之间的最佳平衡点。 Domestic machine translation research began in 1956. In the second year, the Institute of Linguistics, Chinese Academy of Sciences and the Institute of Computer Technology jointly carried out russian-Chinese machine translation research, and the research on post-translation editing mainly focused on the design and development of post-translation editing tools. Huang Heyan (1994) proposed the design principle and algorithm of a new intelligent post-translation editor to improve the efficiency of post-translation editing. Wei Changhong (2007) pointed out the various components of machine translation, post-editing can help translators improve the quality of translation and the efficiency of manual translation, and analyzed the basic concepts of post-editing, the necessity of post-editing, the means of post-editing, the errors that need to be corrected after editing, and the practitioners of post-editing. In 2008, Ge-CCT designed the ge-CCT collaborative Translation Platform (GE-CCT), which combines machine translation with post-translation editing and puts it into practice in the field of business translation. Li Mei (2013) carried out machine translation secondary processing of regular and typical translation errors in parallel corpus of Chinese-English machine translation for automotive majors, namely, automatic filtering of these errors through post-translation editing to improve machine translation speed and machine translation quality. Cui Qiliang (2014) discussed the concept of post-editing, analyzed the application and research status of post-editing, the driving force for the development of post-editing, the subjects suitable for post-editing research, and put forward the code of conduct to improve the quality and efficiency of post-editing. Feng Quangong (2015) discussed the cultivation methods of post-editing from the aspects of the industry demand for post-editing, curriculum setting, editing ability, teaching and tool selection, and pointed out that offering post-editing courses in universities can enhance students' professional competitiveness and better meet the demand of post-editing in the language service industry. Cui Qiliang (2015) analyzed various types of errors in machine translation, including over-translation, under-translation, term translation, form, format, multiple translation and omission, redundancy, part of speech judgment, clause translation, phrase sequence and original sentence structure constraint, and summarized the characteristics of post-translation editing. Feng Quangong (2016) analyzed the current focus of post-translation editing research and pointed out the development trend of future research, such as the development of integrated translation environment, research and development of specific machine translation system, cooperation with post-translation editing, application personnel training and in-depth cooperation at the industry-university-research level. Feng Quangong (2018) attempted to construct a model of post-translation editing ability from three dimensions of cognition, knowledge and skills, and discussed the composition, representation and correlation of each dimension as well as the enlightenment of this model to post-translation editing teaching. Wang Xiangling (2018) by means of "translation studies bibliography" and "machine translation archives" two big database editing process after translation corpus analysis and product evaluation, the factors affecting efficiency, tools, and the research process of personnel training and research methods, explore its research trend, for related research and translation talent training and discipline construction provides a new research perspective and research methods. Wang Xiangling (2019) compared the subjects' differences in translation speed, translation quality and translator's attitude between manual translation and machine translation through keystroke records and questionnaire survey, so as to guide the training of post-translation editing talents in machine translation in the future. Lu Qiang (2019) studied the formulation of general post-editing quality specifications to guide post-editing projects in machine translation based on translation accuracy, completeness and terminology consistency, and also pointed out that the project quality specifications can be adjusted according to specific project characteristics to better guide post-editing work. Zhu Huifen (2020), based on Newmark translation theory, analyzed the errors in Google English translation of the "eight-eight Strategy", explored the translator's post-translation editing principles in the aspects of text, signification, cohesion and nature, and pointed out the translation principles of implicit information supplement and "name from the source". CAI Qiang (2020) used Google Chinese-English machine translation of abstracts of 200 scientific and technological papers in the CNQI database, and found that the main errors in machine translation included vocabulary, syntax, logic and other errors, and the frequency of errors decreased in descending order. He proposed post-translation editing in terms of vocabulary, syntax, logic and punctuation. It has enriched the research of post-translation editing in machine translation of sci-tech texts. Zhao Tao (2021) points out that people have not distinguished the differences between machine translation and post-translation editing well, and it is necessary to comprehensively understand the relative advantages and inherent disadvantages of machine translation from multiple perspectives. Machine translation and post-editing can be classified according to text types and customer needs to adapt to different use environments. It is also mentioned that in the future, it is necessary to increase the teaching and research work of machine translation and post-editing, and to cultivate and improve the core abilities of machine translation and post-editing from the aspects of knowledge, practice and skills. Li Mei (2021) studied the relationship between post-editing time and translation quality of original text by analyzing the results of post-editing test for professional translators in English-Chinese machine translation, so as to help people understand human-computer collaboration, understand the relationship between original text and post-editing, and improve the effectiveness of post-editing. Chen she (2021) through the contrastive analysis of Chinese petroleum geological literature translation in seven major online translation platform at home and abroad, found the main problem is focused on the meaning, part of speech, word order, sentence structure, pausing, noun markers, collocation, punctuation, grammar letter case, information integrity and so on, is put forward based on the principle of "cinda cut" to carry out the translation after editing, To find the best balance between translation processing speed and translation accuracy. Laurian（1984）在讨论机器翻译文本是否需要译后编辑时，指出如果文本中的信息要求人去感知，那么这类文本翻译译文的可信度和准确度就会有所缺失，不适合进行机器翻译，需要进行译后编辑。Reiss（1989）根据功能语言学家K. Bühler的语言功能三分法，将文本分成信息型、表情型和操作型三种类型，并说明文本类型与翻译策略的整体关系。Chesterman（1989）指出表情型文本是一种“创作类产品”，兼具美学特征，信息发出者可以围绕主题自由创造，并有意识地“使用语言的表情和联想意义”，将个人的情感、情绪、态度通过“创造性作品”及对事实“艺术性塑造”加以表达。研究还指出诗歌是最具表情功能的文本，在翻译过程中也是译者参与“再创造”的过程。 Laurian (1984), when discussing whether machine-translated texts need post-translation editing, pointed out that if the information in the texts requires human perception, the credibility and accuracy of the translated texts of such texts will be deficient, which is not suitable for machine translation and post-translation editing is required. Reiss (1989) based on functional linguist K. Buhler's tripartite method of language functions divides text into three types: information type, expression type and operation type, and explains the overall relationship between text type and translation strategy. Chesterman (1989) points out that the expression type of text is a kind of "creative class" products, both aesthetic characteristics, message originators can free created around the theme, and consciously "the use of language expression and meaning", the personal emotion, mood, attitude through the works of the "creative" and "artistic shape" to express the facts. The study also points out that poetry is the most expressive text and the translator participates in the process of "re-creation" in the process of translation. 卓键滨（2018）对比三种不同文本类型采用人工翻译和机器翻译（谷歌翻译）的译文质量和准确度差异，发现机器翻译最擅长处理信息型文本，其次是表达型文本中的非文学翻译，并指出机器翻译必须与人工翻译合作，才能共同促进彼此和翻译事业的发展而不会被时代淘汰。2018年，通过实验法对比韩国文学作品人工翻译与机器翻译译后编辑质量，发现人工翻译所需时长是译后编辑的3倍，译文准确度与流利性都不如译后编辑，指出人工翻译与译后编辑译文产生类似的错误类型，如误译、省略和语法错误等。熊璨（2020）从语言、文化和话语三个层面分析人工智能翻译文学作品和非文学作品的可行性和存在的问题，指出文化和话语层面存在发展障碍，文学类作品单纯依靠人工智能翻译是无法实现的，只有采用人机结合的机器翻译译后编辑模式才能保证人工智能翻译文学作品的质量。翁义明（2020）运用语料库研究方法，对比文学作品中汉语流水句人工翻译和机器翻译英译文的句法和语篇，发现二者在目标语小句主语、动词类型、句子类别、关联词语与译文语序存在巨大差异，并指出机器翻译需要丰富译文动词类型、句法结构和语篇内关联方式，缩小与人工翻译的差距，达到增强文学翻译语言的生动性、地道性和欣赏性。 Zhuo Jianbin (2018) compared the translation quality and accuracy differences between human translation and machine translation (Google Translate) in three different text types, and found that machine translation is best at processing informational text, followed by non-literary translation in expressive text, and pointed out that machine translation must cooperate with human translation. Only in this way can we promote the development of each other and the cause of translation and not be eliminated by The Times. In 2018, through experiment contrast Korean literature human translation and machine translation after editing quality, found after artificial translation required length is 3 times of editing, translation accuracy and fluency are behind after editing, pointed out that after the artificial translation and editing a similar error type, such as mistranslation, ellipsis and grammatical errors. Xiong Can (2020) analyzed the feasibility and existing problems of ai translation of literary and non-literary works from three aspects of language, culture and discourse, and pointed out that there were development barriers in culture and discourse, and literary works could not be translated solely by artificial intelligence. The quality of literary works translated by artificial intelligence can only be guaranteed by using machine translation and post-editing mode combined with human-machine translation. WengYiMing corpus (2020) research methods, comparative literature in Chinese run-on sentence human translation and English translation machine translation, syntactic and discourse found both in the target language clause subject, verb types, sentence category, associated disparity in the word and the word order, and points out that translation machine translation needs to enrich verb types, syntactic structure and text inside connection way, To narrow the gap with artificial translation and enhance the vividness, idiomatic and appreciation of literary translation language.

1.1.3机器翻译以及译后编辑的问题研究 1.1.3 Research on machine translation and post-translation editing 众所周知，信息技术的发展和应用给翻译带来了深刻的影响与变革（孔令然、崔亮，2018；司显柱、郭小洁，2018）。近年来，机器翻译+译前/译后编辑模式成为机器翻译领域的一项重要变革，已经广泛应用在人们的工作和生活中，成为未来翻译工作的一种重要模式。人们对于译后编辑的研究也呈现井喷态势，包括理论探讨，如译后编辑的概念、发展现状、人才培养、能力建设及未来趋势等（魏长宏、张春柏，2007；崔启亮，2014；冯全功、崔启亮，2016；冯全功、刘明，2018），还包括各种实证研究，研究者往往从某个译后编辑实践出发，探讨译后编辑的方法（陈齐祖，2014）和作用（王萍，2016；郭高攀、王宗英，2017）等，同时越来越多年轻的硕士研究生会选择机器翻译的译后编辑作为硕士论文研究主题。目前人们对于机器翻译背后的机制了解不够，因此未能提出具有针对性的译后编辑策略，且现有的机器翻译策略和方法大都停留在句子层面，未涉及段落和篇章翻译的策略和方法，还有待提高。 As we all know, the development and application of information technology has brought profound influence and reform to translation (Kong Lingran, Cui Liang, 2018; Si Xianzhu, Guo Xiaojie, 2018). In recent years, machine translation + pre-translation/post-translation editing mode has become an important revolution in the field of machine translation. It has been widely used in people's work and life and has become an important mode of translation work in the future. People's research on post-translation editing also showed a trend of explosion, including theoretical discussions, such as the concept of post-translation editing, development status, talent training, capacity building and future trends (Wei Changhong, Zhang Chunbai, 2007; Cui Qiliang, 2014; Feng Quangong, Cui Qiliang, 2016; Feng Quangong, Liu Ming, 2018), including various empirical studies, researchers often start from the practice of post-translation editing to discuss the methods of post-translation editing (Chen Qizu, 2014) and the role of post-translation editing (Wang Ping, 2016; Guo Gaopan, Wang Zongying, 2017). Meanwhile, more and more young postgraduate students will choose post-translation editing of machine translation as the research topic of their master's thesis. At present, people do not know enough about the mechanism behind machine translation, so they fail to put forward targeted post-translation editing strategies. Moreover, the existing strategies and methods of machine translation mostly stay at the sentence level, and do not involve the strategies and methods of paragraph and text translation, which need to be improved. 2012年，学者罗季美、李梅通过对比分析机器译文和人工译文，发现前者在词汇层面的错误率高达84.13%，在句法层面的错误也占到总句数的42.45%。2013年，百度开始研究神经网络机器翻译，2015年率先在机器翻译系统中采用深度神经网络。2014年学者杨南在《基于神经网络学习的统计机器翻译研究》中指出基于规则的机器翻译系统难以有效地利用新的资源自动提高翻译系统的性能，而基于统计的机器翻译系统的稳定性和拓展性明显优于其他两种方法，能够自然地处理语言的歧义性，从现有语料库中快速构建高性能的翻译系统，并在增加语料的同时能够自动提升翻译性能。2016年，谷歌发布神经网络机器翻译（GNMT），宣称该系统将翻译质量提高到了接近人工翻译的水平。2018年学者徐雪惠在《机器翻译汉英质量评价》中指出目前的技术还不成熟，不能处理复杂的句子结构或理解文字的深层次含义，并且受到的语料训练比较有限，同时语料库容量不足，难以形成较为系统、完善的训练结果。机器与语言学研究的结合尚不完全，无法从语言学层面科学地处理翻译，比如语篇层面来看机器翻译的句子是不连贯的，其对于文化负载信息和比喻的处理也不尽如人意。 In 2012, scholars Luo Jimei and Li Mei made a comparative analysis of machine translation and artificial translation, and found that the former made up 84.13% of lexical errors and 42.45% of syntactic errors in total sentences. In 2013, Baidu started researching neural network machine translation, and in 2015, it was the first to adopt deep neural network in its machine translation system. Yang Na 2014 scholars in the study of statistical machine translation research based on neural network, "said in a machine translation system based on rules is difficult to effectively use new resources to improve the performance of the translation system automatically, based on statistical machine translation system stability and expansibility is superior to other two methods, can deal with the ambiguity of language, naturally A high performance translation system can be rapidly constructed from the existing corpus, and the translation performance can be automatically improved while the corpus is added. In 2016, Google released Neural Network Machine Translation (GNMT), claiming that the system had improved the quality of translation to a level close to that of human translation. In 2018, xu Xuehui, a scholar, pointed out in Chinese-English Quality Evaluation of Machine Translation that the current technology is not mature enough to process complex sentence structures or understand the deep meaning of words, and the corpus training is relatively limited. Meanwhile, the corpus capacity is insufficient, which makes it difficult to form systematic and perfect training results. The combination of machine and linguistic research is not complete, so it cannot scientifically process translation from the linguistic level. For example, from the textual level, the sentences of machine translation are incoherent, and the processing of culture-loaded information and metaphor is not satisfactory. 2022年3月学者董振恒、任维平、游新冬、吕学强在《融入新能源领域术语知识的机器翻译方法》一文中指出已有的神经机器翻译模型在通用领域取得了较高的翻译质量，但对于涉及专业术语的特定领域，其翻译结果存在着大量漏译、错译现象，术语翻译仍存在较大提升空间。 Scholars Dong Zhenheng, Ren Weiping in March 2022, new winter swimming, Lv Xueqiang in terms of knowledge "into the new energy field machine translation method, the article points out that the existing neural machine translation in the general field obtained higher translation quality, but for specific areas involved in professional terms, the translation results there are a large number of leakage, mistranslation phenomenon, There is still much room for improvement in terminology translation. 2022年4月学者张超轶、陈媛、张聚伟在《融合术语信息的神经机器翻译参数初始化研究》一文中提到电气工程领域英汉机器翻译平行语料稀缺。由于电气领域资源低，有限的双语语料限制了此项量对词本身所包含信息的学习，因而电气工程领域的机器翻译译文准确率也不高。 In April 2022, scholars Zhang Chaoyi, Chen Yuan and Zhang Juwei mentioned the scarcity of parallel corpus for English-Chinese machine translation in the field of electrical engineering in the paper "Research on Parameter Initialization of Neuromachine Translation Integrating Terminology Information". Due to the low resources in the field of electrical engineering and the limited bilingual corpus, the accuracy of machine translation in the field of electrical engineering is not high. 2022年9月学者黄永中在《翻译转换理论视角下机器翻译与人工翻译的对比分析》中提到机器翻译可以胜任那些对于表达形式、风格和文化内涵、意境不做特定要求的翻译工作，而对于散发着强烈艺术个性、独特人文风格与魅力的文本，比如小说、诗歌、散文等文艺作品还不能够胜任。 In September 2022 scholars yong-zhong huang in the perspective of translation theory analysis in machine translation and human translation "mentioned in machine translation can be competent for the for expression form, style and cultural connotation, the specific requirements of the artistic conception is not to do translation work, and for sending out the strong artistic personality, unique cultural style and charm of the text, For example, novels, poems, prose and other literary works are not competent. 研究者郭望皓在《神经机器翻译译文测评及译后编辑研究》中提到，机器翻译错误类型主要分为拼写错误、词汇短语错误、句子句法错误、语义错误四大类；在进行中英翻译时，主要是对词汇与句子进行研究，发现常常出现词汇层面的词性误译，并且没有规律可循，即使同一神经机器翻译在翻译不同句子的同一短语结构时，也会出现不同的译法。因此通过对文学文本进行译后编辑，则可以帮助机器翻译提供更大、更准确的语料资源。其次，由于神经机器翻译模型中缺乏“硬对齐”模块，导致译文中经常会出现漏译，即“该翻译的没有翻译”，最后造成意义的不完整。 Guo Wanghao, a researcher, mentioned in his research on Translation Evaluation and Post-translation Editing in Neural machine Translation that machine translation errors are mainly divided into four categories: spelling errors, vocabulary and phrase errors, sentence syntax errors and semantic errors. When translating from Chinese to English, we mainly study words and sentences, and find that pos mistranslations often occur at the lexical level, and there are no rules to follow. Even when translating the same phrase structure in different sentences with the same neural machine translation, there will be different translation methods. Therefore, post-translation editing of literary texts can help machine translation provide larger and more accurate corpus resources. Secondly, due to the lack of hard-aligned modules in the NEURAL machine translation model, translation omissions often occur in the translation, that is, "the translation is not translated", resulting in incomplete meaning. 基于以上学者对于机器翻译的研究，尽管机器翻译能给我们带来极大的便利，有效处理较多层次较低的翻译任务，但神经网络机器翻译仍存在着诸多问题，包括：1. 机器翻译在某些特定领域受现有语料质量的影响，导致译文准确率不高，如电气领域、新能源领域等；2. 在通用领域下，机器翻译在处理结构复杂的句子以及句子的深层含义还不成熟；3. 机器翻译在处理小说、诗歌、散文等创造性程度高的文本或表现性文本时，由于这些文本重视作者或人物形象的情感表现，词汇语义表达往往不稳定，较为模糊，而机器翻译受自身先天不足的限制，只能表达浅层的语言情境，无法表达深层语言情境、语篇语境和人际语境，因此我们需要提升机器翻译对于不同类别文本的兼容性。 Based on the above researches on machine translation, although machine translation can bring us great convenience and effectively handle many lower-level translation tasks, there are still many problems in neural network machine translation, including: 1. Machine translation is affected by the quality of existing corpus in some specific fields, leading to low accuracy of translation, such as electrical field, new energy field, etc. 2. In the general field, machine translation is still immature in dealing with complex sentences and their deep meanings; 3. When machine translation is dealing with highly creative texts or expressive texts such as novels, poems and essays, the semantic expression of vocabulary is often unstable and fuzzy because these texts attach importance to the emotional expression of the author or the image of the characters. However, machine translation is limited by its inherent deficiencies and can only express shallow linguistic situations. Therefore, we need to improve the compatibility of machine translation for different types of texts.

1.1研究价值 1.2Research value 根据Chesterman对文本类型和翻译的研究，我们了解到文学作品等表情功能文本单纯依靠机器翻译无法实现“艺术再创造”，因而必须依靠人工参与机器翻译译后编辑模式，进行“艺术再创造”才能保证翻译质量。 According to Chesterman's research on text types and translation, we know that emotion-functional texts such as literary works cannot be "recreated artistically" by machine translation alone. Therefore, human participation in post-translation editing mode of machine translation must be relied on for "recreated artistically" to ensure translation quality. 1.提高机器翻译的质量 2.Improve the quality of machine translation

   减少现有机器翻译具有的误译、错译、漏译以及在长、难句处理时出现的句法不通顺、佶屈聱牙译文的数量。
   Reduce the number of mistranslations, mistranslations, omissions and syntactically awkward and suffocating translations that are difficult and difficult in existing machine translation.

3.减少甚至最终取缔机器翻译中的人工参与程度 4.Reduce and eventually eliminate human involvement in machine translation

   大大降低译后编辑人工脑力与体力符合，解放人类的双手和大脑。
   Greatly reduce the post-translation editing artificial brain and physical accord with the liberation of human hands and brain.

5.加速不同语际间的文化传播 6.Accelerate cultural communication between different languages

   机器翻译处理速度远远大于人类，大概5-6倍，且不眠不休，成本低。
   Machine translation processing speed is much faster than human, about five to six times, and never sleeps, low cost.

7.完善机器翻译软件开发 8.Perfect machine translation software development 提高语料库现有文本与待译文本的匹配度，精确搜索算法、扩建双语平行对齐语料库、充实不同文本类型的语料库资源。 Improve the matching degree of the existing texts in the corpus and the texts to be translated, accurately search algorithms, expand the bilingual parallel alignment corpus, and enrich the corpus resources of different text types. 9.促进AI智能的发展 10.Promote the development of AI intelligence 使计算机在具有抽象思维能力的基础上，开发与人类媲美的形象思维能力。 On the basis of abstract thinking ability, computer can develop image thinking ability comparable with human beings. 本研究旨在通过**软件对《鲁迅全集》（共计700万字）实施中德、中英机器翻译，再进行译后编辑，总结机器翻译常见错误类型以及译后编辑处理时常见的问题，并系统总结机器翻译产生错误的原因、采用的策略和方法，继而探讨采取什么策略和方法可以提高机器翻译的准确度？使用什么方法能够减少译后编辑的人工参与度？如何构建适用于任意语对转换的翻译模型？此外，还可以协助工程师完善以自然语言理解为核心的认知计算理论和方法（如算法的创新与组合），提高计算力来推进深度学习（文本学习与深度学习结合），在国内推动中华文化传播与中华典籍外译工作，提升我国人工智能的国际竞争力，在国际上加速不同语际间的文化传播，实现新时期机器翻译发展的大跃进。在此过程中，将通过为期2年左右的时间，创建高质量的中德、中英双语平行语料库（《鲁迅全集》为文本，语料库总字数约1400万字），采用语料库语言学法、跨学科研究法和数量研究法等研究方法。 This study aims to implement machine translation from Chinese to German and Chinese to English for the Complete Works of Lu Xun (7 million words in total) by ** software, and then conduct post-translation editing, summarize the common types of errors in machine translation and the common problems in post-translation editing, and systematically summarize the causes of errors in machine translation, strategies and methods adopted. Then, what strategies and methods can be adopted to improve the accuracy of machine translation? What can be done to reduce human involvement in post-translation editing? How to construct a translation model for any Italian pair conversion? In addition, can also help engineers improve the natural language understanding is the core of cognitive computational theory and methods, such as innovation and composition) of the algorithm, improve the computing power to promote deep learning (text combine learning and deep learning), outside the domestic promoting the dissemination of Chinese culture and the Chinese classics translation work, enhance the international competitiveness of China's artificial intelligence, It will accelerate the cultural communication between different languages internationally and achieve a great leap forward in the development of machine translation in the new era. In this process, a high-quality Bilingual parallel corpus of German and Chinese and English will be created in about two years (the Complete Works of Lu Xun is the text, with a total of about 14 million words in the corpus), and research methods such as corpus linguistic method, interdisciplinary research method and quantitative research method will be adopted.

2. 研究内容（本课题的研究对象、总体框架、重点难点、主要目标等） 2. Research content (research object, overall framework, key and difficult points, main objectives, etc.) 2.1研究对象 2.1 Research Objects 研究对象包括：《鲁迅全集》的中、德、英多语平行语料库（常见错误类型、原因和策略）；译后编辑人工参与度；适用于任意语对转换的翻译模型；认知计算理论和方法、计算力和深度学习。 The research objects include: the Chinese, German and English parallel corpus of the Complete Works of Lu Xun (common error types, causes and strategies); Human participation in post-translation editing; A translation model applicable to any Italian translation; Cognitive computing theory and methods, computational power and deep learning. ， . 2.2 总体框架 2.2 General Framework I. 文献综述 I. Literature review II. 机器翻译 II. Machine translation II.I 机器翻译的定义 II.I Definition of machine translation II.II 机器翻译发展的历史 II.II History of machine translation development II.III 机器翻译的特征 II.III Features of machine translation II.IV 机器翻译的不足 II.IV Shortcomings of machine translation III. 译后编辑 III. Post-translation and editing

 III.I 译后编辑的定义
 III.I Definition of post-translation editing
 III.II译后编辑发展的历史
 III.II History of post-translation editing
 III.III 译后编辑的特征
 III. Characteristics of post-translation editing
 III.IV 译后编辑的不足
 Iii. shortcomings of post-translation editing in IV

IV. 《鲁迅全集》中、德双语平行语料库 IV. Chinese and German bilingual parallel corpus of The Complete Works of Lu Xun

 IV.I 《鲁迅全集》中、德双语平行语料库文本对应之处
 Iv. I The correspondence between Chinese and German bilingual parallel corpus texts in The Complete Works of Lu Xun
      IV.I.I 对应的类型
      Iv.i.i Corresponding type
      IV.I.II 对应的原因
      IV.i.i
      IV.I.III 建议
      IV. I.I II
 IV.II 《鲁迅全集》中、德双语平行语料库文本不对应之处
 IV.II The inconsistencies between Chinese and German bilingual parallel corpus texts in The Complete Works of Lu Xun
      IV.I.I 不对应的类型
      Iv.i.i Does not correspond to the type
      IV.I.II 不对应的原因
      Iv.i.i. Causes of the mismatch
      IV.I.III 建议
      IV. I.I II

V. 《鲁迅全集》中、英双语平行语料库 V. Chinese and English bilingual corpus of The Complete Works of Lu Xun

 V.I 《鲁迅全集》中、英双语平行语料库文本对应之处
 V. The correspondence between English and Chinese bilingual parallel corpus texts in The Complete Works of Lu Xun
      V.I.I 对应的类型
      The type of V.I.I
      V.I.II 对应的原因
      V.I.I
      V.I.III 建议
      V.I.I II advice
 V.II 《鲁迅全集》中、英双语平行语料库文本不对应之处
 V.I., the Complete Works of Lu Xun, and the English bilingual parallel corpus texts do not correspond
      V.I.I 不对应的类型
      V.I.I does not correspond to a type
      V.I.II 不对应的原因
      The reason why V.I.I does not correspond
      V.I.III 建议
      V.I.I II advice

VI. 研究结果实际应用 VI. Practical application of research results VI.I 机器翻译自动译后编辑器开发 VI.I Machine translation automatic post-translation editor development VI.II 任意语对转换的翻译模型开发 VI.II Development of translation model for any Italian translation VI.III 认知计算理论和方法升级（如算法的创新与组合） VI.III Upgrading of cognitive computing theory and methods (e.g. innovation and combination of algorithms) VI.IV 提高计算力以推进深度学习能力（文本学习与深度学习结合） VI.IV Improving computing power to promote deep learning ability (combination of text learning and deep learning) VII. 研究意义 VII. Research significance VII.I 提升国家及公司人工智能领域的国际竞争力 Vii.i Enhance national and corporate international competitiveness in the field of ARTIFICIAL intelligence VII.II 加速不同语际间的文化传播 Vii.ii Accelerates cultural transmission between languages VII.III 提高机器翻译的处理速度和质量 Vii.iii Improve the processing speed and quality of machine translation VII.IV 减少译后编辑的人工参与度 Vii.iv Reduces human participation in post-translation editing VII.V 推动中华文化传播与中华典籍外译工作 Vii.v to promote the dissemination of Chinese culture and the translation of Chinese classics

2.3重点难点 2.3 Key points and Difficulties 2.3.1研究重点 2.3.1 Research Focus 采用合理的机器翻译和译后编辑完成《鲁迅全集》英、德多语平行语料库建设，收集机器翻译和译后编辑过程中出现的各种错误和不完善的地方。 The construction of a multi-language parallel corpus of The Complete Works of Lu Xun is completed with reasonable machine translation and post-translation editing, and various errors and imperfections in the process of machine translation and post-translation editing are collected. 2.3.2研究难点 2.3.2 Research difficulties （1）分析《鲁迅全集》平行语料库中各种错误出现的原因和解决的策略 （2）This paper analyzes the causes and solutions of various errors in the parallel corpus of The Complete Works of Lu Xun （3）如何减少译后编辑的人工参与度 （4）How to reduce human involvement in post-translation editing （5）如何协助工程师完善软件系统（适用于任意语对转换的翻译模型、算法的创新与组合、提高计算力来推进深度学习、） （6）How to assist engineers to improve software systems (translation models for arbitrary translation, innovation and combination of algorithms, improvement of computing power to promote deep learning,

2.4主要目标 2.4 Main Objectives （1）提高机器翻译的质量 （2）Improve the quality of machine translation （3）减少译后编辑的人工参与度 （4）Reduce human involvement in post-editing （5）提高计算力来推进深度学习 （6）Improve computing power to advance deep learning 3. 思路方法（本课题研究的基本思路、具体研究方法、研究计划及其可行性等） 3. Ideas and methods (basic ideas, specific research methods, research plan and feasibility of this research, etc.) 3.1基本思路 3.1 Basic Ideas

   梳理现有《鲁迅全集》和译本，了解**机器翻译软件的特点以及机器翻译译后编辑的方法、策略和原则的基础上，结合文本类型分类说，建立《鲁迅全集》多语平行语料库，分析译文错误类型并加以修正，总结错误规律和常规解决办法，协助工程师完善软件系统，包括适用于任意语对转换的翻译模型、算法的创新与组合、提高计算力来推进深度学习等，提升机器翻译的质量，减少译后编辑的人工参与度，解放人类的双手和大脑，在国内推动中华文化传播与中华典籍外译工作，提升我国人工智能的国际竞争力，在国际上加速不同语际间的文化传播，实现新时期机器翻译发展的大跃进。。
   Combing the existing "lu xun complete works" and translation, understanding of the features of * * machine translation software and translation machine translation after editing methods, strategies and principles on the basis of the combination of text type classification, establish a "lu xun complete works" multilingual parallel corpus, analyzing the error types and modified, summarizes the law of error and conventional solutions, assist engineer to improve the software system, Including applicable to the former Italian translation model of transformation, innovation and combination algorithm, improve the computing power to promote deep learning, improve the quality of machine translation, reduce editor after artificial participation, the liberation of human hands and brain, outside the domestic promoting the dissemination of Chinese culture and the Chinese classics translation work, enhance the international competitiveness of China's artificial intelligence, It will accelerate the cultural communication between different languages internationally and achieve a great leap forward in the development of machine translation in the new era. .

3.2研究方法 3.2 Research methods 语料库语言学法、跨学科研究法和数量研究法等研究方法 Corpus linguistic approach, interdisciplinary approach, quantitative approach and other research methods 3.3研究计划 3.3 Research Plan 2022年3月 - 2022年7月 《鲁迅全集》语料输入 March 2022 -- July 2022 Corpus input of The Complete Works of Lu Xun 2022年8月 - 2023年8月 《鲁迅全集》德、英文机器翻译及译后编辑处理 Aug. 2022 -- Aug. 2023: German and English machine translation and post-translation editing of The Complete Works of Lu Xun 2023年9月 - 2023年12月 对《鲁迅全集》英文、德文译本校勘 From September 2023 to December 2023, collated the English and German versions of the Complete Works of Lu Xun 2023年12月 - 2024年3月 译文错误类型归类并修正；总结错误规律和常规解 From December 2023 to March 2024, classification and correction of translation errors; Summarize error laws and conventional solutions 决办法；可协助工程师完善软件系统（适用于任意语对转换的翻译模型、算法的创新与组合、提高计算力来推进深度学习、） Definitely way; It can assist engineers to improve the software system (suitable for translation models of arbitrary language pairs, innovation and combination of algorithms, improvement of computing power to promote deep learning, 3.4可行性 3.4 the feasibility （1）符合国家中华文化“走出去”战略方针，有利于推动中国文化继承、发展、延续和传播，使中华文化永葆青春。 (1) In line with the national Strategy of "going global" of Chinese culture, it is conducive to promoting the inheritance, development, continuation and dissemination of Chinese culture and keeping Chinese culture young forever. （2）项目主持人精通中、德语、英语，资深汉学家，翻译了曹雪芹的《红楼梦》及鲁迅、周作人、许地山、郁达夫、朱自清、冰心、巴金、钱钟书、贾平凹等众多著名国内作家的经典著作。 (2) The project host is proficient in Chinese, German and English, and is a senior Sinologist. He has translated Cao Xueqin's Dream of Red Mansions and the classic works of lu Xun, Zhou Zuoren, Xu Dishan, Yu Dafu, Zhu Ziqing, Bing Xin, Ba Jin, Qian Zhongshu, Jia Pingwa and many other famous Chinese writers. （3）项目执行者包括MTI及翻译方向的教师、研究生，具有合格的资质和能力开展《鲁迅全集》英译和德译机器翻译译后编辑工作。 (3) Project implementers include teachers and postgraduates from MTI and translation, who are qualified and capable of translating the Complete Works of Lu Xun into English and German machine translation. （4）项目与课程内容（中华典籍外译）紧密结合，理论联系实际，线上线下结合。 (4) The project is closely combined with the course content (Translation of Chinese classics), theory is connected with practice, and online and offline are combined. （5）**翻译软件公司的智力支持和资助。 (5) Intellectual support and funding from ** translation software Company. 4. 创新之处（在学术思想、学术观点、研究方法等方面的特色和创新） 4. Innovation (characteristics and innovation in academic thoughts, viewpoints, research methods, etc.) （1）扩充机器翻译语料库建设文本类型（表情型文本）和规模（1400万字） (1) Expand the text type (emoticons) and size (14 million words) of machine translation corpus construction （2）减少甚至取缔译后编辑的人工参与度 (2) Reduce or even ban the manual participation of post-translation editors 5. 预期成果（成果形式、使用去向及预期社会效益等） 5. Expected results (results form, use direction and expected social benefits, etc.) （1）创建《鲁迅全集》中、德、英多语平行语料库 (1) Create a multi-language parallel corpus in Chinese, German and English for the Complete Works of Lu Xun （2）开发高质量机器翻译自动化编辑软件 (2) Develop high-quality automatic editing software for machine translation （3）构建适用于任意语对转换的翻译模型，开发国际语翻译模型 (3) Construct translation models suitable for any Italian translation and develop interlingua translation models （4） 协助工程师完善以自然语言理解为核心的认知计算理论和方法（如算法的创新与组合）、提高计算力来推进深度学习（文本学习与深度学习结合） (4) Assist engineers to improve cognitive computing theories and methods centered on natural language understanding (such as innovation and combination of algorithms) and improve computing power to promote deep learning (combination of text learning and deep learning) 5.1成果形式与使用去向 5.1 Form and use of results （1）创建《鲁迅全集》中、德、英多语平行语料库，填补此前空白，丰富公司现有语料库规模和文本类型。 (1) Create a multi-language parallel corpus in Chinese, German and English for the Complete Works of Lu Xun to fill the previous gaps and enrich the company's existing corpus scale and text types. （2）总结机器翻译和译后编辑存在的问题和解决的策略，开发高质量机器翻译自动化编辑软件，提高机器翻译的处理速度和效率，减少译后编辑的人工参与度，解放人类的双手和大脑。 (2) Summarize the problems existing in machine translation and post-translation editing and the strategies to solve them, develop high-quality automatic editing software for machine translation, improve the processing speed and efficiency of machine translation, reduce the human participation in post-translation editing, and liberate human hands and brain. （3）构建适用于任意语对转换的翻译模型，开发国际语翻译模型，减少语际交流困难、产生的误解，增进不同语种人际交流效果。 (3) Construct translation models suitable for any Italian pair conversion, develop interlanguage translation models, reduce inter-language communication difficulties and misunderstandings, and improve interpersonal communication in different languages. （4） 协助工程师完善以自然语言理解为核心的认知计算理论和方法（如算法的创新与组合）、提高计算力来推进深度学习（文本学习与深度学习结合）来提高机器翻译的反应速度、准确度、匹配度，从而提高机器翻译和译后编辑的质量。 (4) assist engineer to improve the natural language understanding is the core of cognitive computational theory and methods, such as innovation and composition) of the algorithm and improve the computing power to promote deep learning (text combine learning and deep learning) to improve the response speed and accuracy of machine translation, matching degrees, so as to improve the quality of machine translation and translation editing. 5.2 预期效益 5.2 Expected Benefits （1）《鲁迅全集》中、德、英多语平行语料库，所有权在项目主持人，公司资助方可以留存复印件。 (1) The multi-language parallel corpus of The Complete Works of Lu Xun in German and English is owned by the project host, and the company sponsor can keep copies. （2）开发的高质量机器翻译自动化编辑软件、归公司所有，但项目主持人和团队单位可以享有免费使用权。 (2) The high-quality automatic editing software of machine translation developed is owned by the company, but the project host and team unit can enjoy the right to use it free of charge. （3）任意语对转换的翻译模型、认知计算理论和方法、提高计算力来推进深度学习等软件方面的开发和设计归公司所有，但项目主持人和团队单位可以享有附带效应。 (3) The development and design of software such as translation model of any Italian translation, theory and method of cognitive computing, and improvement of computing power to promote deep learning shall be owned by the company, but the project host and team unit can enjoy side effects. 6. 参考文献（开展本课题研究的主要中外参考文献） 6. References (main Chinese and foreign references to carry out this research) Yngve, Victor H. “About mechanical Translation.” Mechanical Translation 1.1 (1954a): 1-2. Yngve, Victor H. "About mechanical Translation." Mechanical Translation 1.1 (1954A): 1-2. Yngve, Victor H. “The machine and the man.” Mechanical Translation 1.2 (1954b): 21-22. Yngve, Victor H. "The machine and The man." Mechanical Translation 1.2 (1954B): 21-22. Miller, George A. and J. G. Beebe-Center. “Some psychological methods for evaluating the quality of translations.” Mechanical Translation 3.3 (1956): 73-80. Miller, George A. and J. G. Beebe-Center. "Some psychological methods for evaluating the quality of translations." Mechanical Translation 3.3 (1956): 73-80. ALPAC. Language and machines: computers in translation and linguistics. A report by the Automatic Language Processing Advisory Committee, Division of Behavioral Sciences, National Academy of Sciences, National Research Council. Washington, D.C.: National Research Council, 1966: 124. ALPAC. Language and machines: computers in translation and linguistics. A report by the Automatic Language Processing Advisory Committee, Division of Behavioral Sciences, National Academy of Sciences, National Research Council. Washington, D.C.: National Research Council, 1966: 124. Orr, David B. and Victor H. Small. “Comprehensibility of machine-aided translations of Russian scientific documents.” Mechanical Translation and Computational Linguistics 10.1-2 (1967): 1-10. Orr, David B. and Victor H. Small. "Comprehensibility of machine- translations of Russian scientific documents." Mechanical Translation and Computational Linguistics 10.1-2 (1967): 1-10. Van Slype, Georges. “Systran: evaluation of the 1978 version of the Systran English - French automatic system of the Commission of the European Communities.” The Incorporated Linguist, 18.3 (1979): 86-89. Van Slype, Georges. "Systran: evaluation of the 1978 version of the Systran English - French automatic system of the Commission of the European Communities." The Incorporated Linguist, 18.3 (1979): 86-89. Green, Roy. “The MT errors which cause most trouble to post-editors.” Practical experience of machine translation. Ed. Veronica Lawson. Amsterdam, New York, Oxford: North-Holland Publishing Company, 1982. 101-104. Green, Roy. "The MT errors which cause most trouble to post-editors." Practical experience of machine translation. Ed. Veronica Lawson. Amsterdam, New York, Oxford: North-Holland Publishing Company, 1982. 101-104. Sereda, Stanley P. “Practical experience of machine translation.” Practical experience of machine translation. Ed. Veronica Lawson. Amsterdam, New York, Oxford: North-Holland Publishing Company, 1982. 119-127. Sereda, Stanley P. "Practical experience of machine translation." Practical experience of machine translation. Ed. Veronica Lawson. Amsterdam, New York, Oxford: North-Holland Publishing Company, 1982. 119-127. Laurian, Anne-Marie. “Machine translation: what type of post-editing on what type of documents for what type of users.” Coling84: 10th International Conference on Computational Linguistics & 22nd Annual Meeting of the Association for Computational Linguistics, 2-6 July 1984. California: Stanford University, 1984. 236-238. Laurian, Anne-Marie. "Machine translation: What type of post-editing on what type of documents for what type of users." Coling84: 10th International Conference on Computational Linguistics & 22nd Annual Meeting of the Association for Computational Linguistics, 2-6 July 1984. California: Stanford University, 1984. 236-238. Wagner, Elizabeth. “Rapid post-editing of Systran”. Tools for the trade: Translating and the Computer 5. Ed. Veronica Lawson. London: Aslib, 1985. Wagner, Elizabeth. "Rapid post - editing of Systran". Tools for the trade: Translating and the Computer 5. Ed. Veronica Lawson. London: Aslib, 1985. Vasconcellos, Muriel and Marjorie Léon. “SPANAM and ENGSPAN: machine translation at the Pan American Health Organization.” Computational Linguistics 11.2-3 (1985): 122-136. Vasconcellos, Muriel and Marjorie Leon. "SPANAM and ENGSPAN: Machine Translation at the Pan American Health Organization." Computational Linguistics 11.2-3 (1985): 122-136. Schneider, Thomas. “The METAL system, status 1989.” MT Summit II, August 16-18, 1989, Munich, Germany, 1989. 128-136. Schneider, Thomas. "The METAL System, Status 1989." MT Summit II, August 16-18, 1989, Munich, Germany, 1989. 128-136. Reiss K. Text Types, Translation Types and Translation Assessment (Tr. by A. Chesterman) [A]. In Chester-man A. (ed.). 1989: 105-115. Reiss K. Text Types, Translation Types and Translation Assessment (Tr. by A. Chesterman) [A]. In Chester-man A. (ed.). 1989:105-115. Chesterman A. ed. Readings in Translation Theory [C]. Helsinki: Oy Finn Lectura Ab, 1989. Chesterman A. ed. Readings in Translation Theory [C]. Helsinki: Oy Finn Lectura Ab, 1989. Allen, J. Post-editing. In: H. Somers (ed.) Computers and Translation: A Translator’s Guide[M]. John Benjamins Publishing, Amsterdam, 2003. Allen, J. Post-editing. In: H. Somers (ed.) Computers and Translation: A Translator's Guide[M]. John Benjamins Publishing, Amsterdam, 2003. Senez, Dorothy. “Post-editing service for machine translation users at the European Commission.” Translating and the Computer 20. Proceedings from Aslib conference, 12 - 13 November 1998. Senez, Dorothy. "Post-editing service for machine translation users at the European Commission." Translating and the Computer 20. Proceedings from Aslib conference, 12 - 13 November 1998. Krings, Hans P. Repairing Texts. Empirical Investigations of Machine Translation Post-Editing Processes. Trans. Geoffrey S. Koby, Gregory M. Shreve, Katja Mischerikow, and Sara Litzer. Kent, Ohio: The Kent State University, 2001. Krings, Hans P. Repairing Texts. Empirical Investigations of Machine Translation Post-Editing Processes. Trans. Geoffrey S. Koby, Gregory M. Shreve, Katja Mischerikow, and Sara Litzer. Kent, Ohio: The Kent State University, 2001. John Nebonne. Computer-Assisted Language Learning and Natural Language Processing. In: Ruslan Mitkov (ed.) Handbook of Computational Linguistics Oxford University Press, 2002: 670-698. John Nebonne. Computer-Assisted Language Learning and Natural Language Processing. In: Ruslan Mitkov (Ed.) Handbook of Computational Linguistics Oxford University Press, 2002: 670-698. Allen, J. Post-editing. In: H. Somers (ed.) Computers and Translation: A Translator’s Guide[M]. John Benjamins Publishing, Amsterdam, 2003. Allen, J. Post-editing. In: H. Somers (ed.) Computers and Translation: A Translator's Guide[M]. John Benjamins Publishing, Amsterdam, 2003. Guerberof, A. Productivity and quality in MT post-editing, In: Proceedings of MT Summit XII-Workshop: Beyond Translation Memories: New Tools for Translators MT, AMTA 2009 (26-30 Aug. 2009, Ottawa, Canada). Guerberof, A. Productivity and quality in MT post-editing, In: Proceedings of MT Summit XII-Workshop: Beyond Translation Memories: New Tools for Translators MT, AMTA 2009 (26-30 Aug. 2009, Ottawa, Canada). Midori Tatsumi. Post-Editing Machine Translated Text in A Commercial Setting: Observation and Statistical Analysis[D]. Ireland: Dublin City University, 2010. Midori Tatsumi. Post-Editing Machine Translated Text in A Commercial Setting: Observation and Statistical Analysis[D]. Ireland: Dublin City University, 2010. O’ Brien, S. Introduction to Post-Editing: Who, What, How and Where to Next? 2010. http://amta2010.amtaweb.org/AMTA/papers/6-01-ObrienPostEdit.pdf (consulted 28. 3. 2022). O 'Brien, S. Introduction to post-editing: Who, What, How and Where to Next? 2010. http://amta2010.amtaweb.org/AMTA/papers/6-01-ObrienPostEdit.pdf (consulted 28. 3. 2022). TAUS. (2010). MT Post-editing Guidelines, available at https://www.taus.net/academy/best-practices/postedit-best-practices/machine-translation-post-editing-guidelines TAUS. (2010). MT Post-editing Guidelines, available at https://www.taus.net/academy/best-practices/postedit-best-practices/machine-translation-post-editing-guidelines Garcia, I. Translating by post-editing: Is it the way forward [J]. Machine Translation, 2011( 3). Garcia, I. Translating by post-editing: Is it the way forward [J]. Machine Translation, 2011( 3). Sharon O’ Brien, Michel Simard and Lucia Specia. The Workshop on Post-editing Technology and Practice of the 14th Machine Translation Summit [C]. Switzerland: The European Association for Machine Translation, 2013. Sharon O 'Brien, Michel Simard and Lucia Specia. The Workshop on Post-editing Technology and Practice of the 14th Machine Translation Summit [C]. Switzerland: The European Association for Machine Translation, 2013. ISO TC 37. ISO 18587 Translation services - Post-editing of machine translation output - Requirements[Z]. 2014. ISO TC 37. ISO 18587 Translation services - Post-editing of machine translation output - Requirements[Z]. 2014. No, Lee Joo. A Study on the different versions of the Complete Works of Lu Xun[J]. Journal of Humanities, 2017: 285-308. No, Lee Joo. A Study on the different versions of the Complete Works of Lu Xun[J]. Journal of Humanities, 2017: 285-308. ISO 18587. Translation services - Post-editing of machine translation output - Requirements. Geneva: International Organization for Standardization. 2017. ISO 18587. Translation services - Post-editing of machine translation output - Requirements. Geneva: International Organization for Standardization. 2017. Koponen, M. & L. Salmi. Post-editing quality: Analysing the correctness and necessity of post-editor corrections [J]. Linguistica Antverpiensia, New Series: Themes in Translation Studies, 2017( 16) . Koponen, M. & L. Salmi. Post-editing quality: Analysing the correctness and necessity of post-editor corrections [J]. Linguistica Antverpiensia, New Series: Themes in Translation Studies, 2017( 16) . Lommel, Arle. “Augmented translation: A new approach to combining human and machine capabilities.” Proceedings of the 13th Conference of the Association for Machine Translation in the Americas (Volume 2: User Papers), Boston, March 17 - 21, 2018. Lommel, Arle. Augmented Translation: A new approach to Combining human and machine capabilities. Proceedings of the 13th Conference of the Association for Machine Translation in the Americas (Volume 2: User Papers), Boston, March 17 - 21, 2018. A Quality Comparison of English Translations of Korean Literature between Human Translation and Post-Editing[J]. International Journal of Advanced Culture Technology, 2018, 6(4): 165-171. A Quality Comparison of English Translations of Korean Literature between Human Translation and Post-Editing[J]. International Journal of Advanced Culture Technology, 2018, 6(4): 165-171. Sakamoto, A. Why do many translators resist post-editing? A sociological analysis using Bourdieu's concepts [J]. The Journal of Specialised Translation, 2019( 31). Sakamoto, A. Why do many translators resist post-editing? A sociological analysis using Bourdieu's concepts [J]. The Journal of Specialised Translation, 2019( 31). Seung HM. Defining language dependent post-editing guide - lines for specific content: The case of the English-Korean pair to improve literature machine translation styles [J]. Babel, 2020, 66(4/5): 811-828. Seung HM. Defining language dependent post-editing guide - lines for specific content: The case of the English-Korean pair to improve literature machine translation styles [J]. Babel, 2020, 66(4/5): 811-828. Zhang, Wen-li. Research on the Translation Strategies and Skills of the Korean Version of “Complete Works of Lu Xun’s Short Stories”[J]. Journal of Chinese Humanities. 2020: 91-101. Zhang, Wen-li. Research on the Translation Strategies and Skills of the Korean Version of "Complete Works of Lu Xun's Short Stories "[J]. Journal of Chinese Humanities. 2020:91-101. Koponen, M. et al. Translation Revision and Post-Editing: Industry Practices and Cognitive Processes [M]. London: Routledge, 2021. Koponen, M. et al. Translation Revision and Post-Editing: Industry Practices and Cognitive Processes [M]. London: Routledge, 2021. 丸山昇, 刘献彪, 刘秉云. 评新版《鲁迅全集》[J]. 山东师范大学学报(人文社会科学版), 1983(3). Maru Shansheng, Liu Xianbiao, Liu Bingyun. Review of the New edition of The Complete Works of Lu Xun [J]. Journal of Shandong Normal University (Humanities and Social Sciences Edition), 1983(3). 江小蕙. 丸山昇评一九八一年版《鲁迅全集》[J]. 鲁迅研究月刊, 1983(4). Jiang Xiaohui. Comments on 1981 edition of Lu Xun's Complete Works by Maruyama Sheng [J]. Lu Xun Studies Monthly, 1983(4). 鲁歌. 对1981年出版的《鲁迅全集》的若干校勘[J]. 绍兴文理学院学报(社科版), 1984(1). Lu Ge. Some Emendations of The Complete Works of Lu Xun published in 1981 [J]. Journal of Shaoxing University of Arts and Sciences (Social Science Edition), 1984(1). 周铭. 嵇康生于何年? —— 对1981年版《鲁迅全集》注释的一点意见[J]. 鲁迅研究月刊, 1984(3). When was Ji Kang born? -- A Comment on the annotation of 1981 edition of The Complete Works of Lu Xun [J]. Lu Xun Studies Monthly, 1984(3). 孙玉石. 中日文化交流史上的又一座丰碑 —— 日本新译二十卷本《鲁迅全集》陆续出版[J]. 鲁迅研究月刊, 1985(7). Sun Yushi. Another monument in the History of Sino-Japanese cultural exchange -- The new Japanese translation of the complete Works of Lu Xun published in Succession [J]. Lu Xun Studies Monthly, 1985(7). 黄河燕, 陈肇雄. 一种智能译后编辑器的设计及其实现算法[J]. 软件学报, 1994(3)． Huanghe Yan, Chen Zhaoxiong. Design and Implementation algorithm of an intelligent post-translation editor [J]. Journal of Software, 1994(3). 武德运. 关于《鲁迅全集》的版本 —— 纪念鲁迅逝世六十周年[J]. 江苏图书馆学报, 1996(6). Wu Deyun. Edition of The Complete Works of Lu Xun -- To commemorate the sixtieth Anniversary of Lu Xun's death [J]. Journal of Jiangsu Library, 1996(6). 张文江. 论《鲁迅全集》的三次编纂及其意义[J]. 上海社会科学院学术季刊, 1996(4): 148-158. Zhang Wenjiang. On the Three Compilations of The Complete Works of Lu Xun and its Significance [J]. Academic Quarterly of Shanghai Academy of Social Sciences, 1996(4): 148-158. 杨知秋.《鲁迅全集》注释补正[J]. 鲁迅研究月刊, 2001(1):70-72. Yang Zhiqiu. Annotation and correction of The Complete Works of Lu Xun [J]. Lu Xun Studies Monthly, 2001(1):70-72. 詹卫东、常宝宝、俞士汶. 机器翻译与语言研究[J]. 语言科学, 2002(1): 47-61. Zhan Weidong, Chang Baobao, Yu Shiwen. Machine Translation and Language Research [J]. Journal of Language Science, 2002(1): 47-61. 魏长宏, 张春柏. 机器翻译的译后编辑[J]. 中国科技翻译, 2007(3). Wei Changhong, Zhang Chunbai. Journal of Machine Translation, 2007(3). 徐斯年.《鲁迅全集》第10卷修订札记[J]. 鲁迅研究月刊, 2007(3): 4-25. [Xu Si-nian. Notes on the Revision of Volume 10 of The Complete Works of Lu Xun [J]. Lu Xun Studies Monthly, 2007(3): 4-25.] 俞凯. 自然语言处理中的认知计算[J]. Interface智能机器系列活动, 2009. Yu K. Cognitive computation in natural language processing [J]. Interface Intelligent Machine Series Event, 2009. 陈丹. 浅析语料库对于翻译研究的意义. 科技视界[J], 2013(34): 22. Chen Dan. On the significance of corpus to translation studies. Science and Technology Vision [J], 2013(34): 22. 李梅, 朱锡明. 译后编辑自动化的英汉机器翻译新探索[J]. 中国翻译, 2013(4). Li Mei, Zhu Ximing. Chinese Translation, 2013(4). 崔启亮. 论机器翻译的译后编辑[J]. 中国翻译, 2014, 35(6): 68-73. On Post-Translation Editing in Machine Translation. China Translation, 2014, 35(6): 68-73. 冯全功, 张慧玉. 全球语言服务行业背景下译后编辑者培养研究[J]. 外语界, 2015(1). Feng Quangong, Zhang Huiyu. Research on the training of post-translation editors in the context of global Language Service Industry [J]. Foreign Languages, 2015(1). 崔启亮. 译后编辑错误类型研究——基于科技文本英汉机器翻译[J]. 中国科技翻译, 2015(4): 19-22. Cui Qiliang. A Study on Post-Translation Editing Errors: Based on ENGLISH-Chinese Machine Translation of Sci-Tech Texts [J]. China Science & Technology Translation, 2015(4): 19-22. 宋绍香. 俄苏鲁迅译介与研究六十年[J]. 鲁迅研究, 2016(3). Song Shaoxiang. Sixty Years of Translation and research on Lu Xun in Russia and The Soviet Union [J]. Lu Xun Studies, 2016(3). 佘晓玲. 鲁迅小说俄译研究述略[J]. 俄罗斯文艺, 2016(1). She Xiaoling. Research on Lu Xun's Novels translated into Russian [J]. Russian Literature and Art, 2016(1). 冯全功, 崔启亮. 译后编辑研究: 焦点透析与发展趋势[J]. 上海翻译, 2016(6). Feng Quangong, Cui Qiliang. Research on Post-translation Editing: Focus on Dialysis and Development Trends [J]. Shanghai Translation, 2016(6). 胡开宝、李毅. 机器翻译特征及其与人工翻译关系的研究. 中国翻译, 2016(5): 10-14. Hu Kaibao, Li Yi. Research on the Characteristics of machine translation and its relationship with human translation. Chinese Translation, 2016(5): 10-14. 屈亚媛、周玉梅. 机器翻译还是人工翻译？ —— 浅析《黄帝内经·素问》双字格养生术语机译错误人工评测[J]. Negative, 2016. (4): 50-53. Qu Yayuan and Zhou Yumei. Machine translation or human translation? [J]. Negative, 2016. (4) : 50-53. 国务院. 国务院关于印发“十三五”国家科技创新规划的通知[Z], 2016-08-08. The State Council. Circular of The State Council on printing and distributing the 13th Five-Year National Science and Technology Innovation Plan [Z], 2016-08-08. 国务院. 2017年国务院政府工作报告[Z], 2017-03-05. The State Council. 2017 Government Work Report of The State Council [Z], 2017-03-05. 国务院. 国务院关于印发新一代人工智能发展规划的通知[Z], 2017-07-08. The State Council. Circular of The State Council on The Issuance of the Development Plan of the New Generation of artificial intelligence [Z], 2017-07-08. 季凯文. 人工智能时代带来的机遇及应对策略[J]．中国国情国力, 2017, 26(12) : 26-28． Ji KAIwen. Opportunities brought by the era of artificial intelligence and countermeasures [J]. China National Conditions and National Strength, 2017, 26(12) : 26-28.] 丁亮、何彦青. 融合领域知识与深度学习的机器翻译领域自适应研究[J]. 情报科学, 2017(10): 125-132. Ding Liang, He Yanqing. Adaptive Research of Machine Translation Domain integrating Domain Knowledge and Deep Learning [J]. Information Science, 2017(10): 125-132.] 冯志伟. 机器翻译与人工智能的平行发展[J]. 外国语, 2018(6)35-48. Feng Zhiwei. Parallel Development of Machine Translation and artificial Intelligence [J]. Foreign Languages, 2018(6)35-48. 冯全功. 译后编辑能力三维模型构建[J]. 外语界, 2018(3): 55-61. Feng Quigong. A Three-dimensional Model of Post-translation Editing Ability [J]. Foreign Language Studies, 2018(3): 55-61. 卓键滨. 机器翻译对各类型文本的德汉翻译能力探究[J].语言文化, 2018(28). Journal of Machine Translation, 2018(28). (in Chinese) 王湘玲, 贾艳芳. 21世纪国外机器翻译译后编辑实证研究[J]. 湖南大学学报(社会科学版), 2018(2). Wang Xiangling, Jia Yanfang. Journal of Hunan University (Social Science Edition), 2018(2). 王湘玲, 王婷婷. 人工翻译与机器翻译译后编辑对比实证研究[J]. 外国语言与文化, 2019(4). Wang Xiangling, Wang Tingting. A Comparative Study on post-translation editing between Human translation and machine translation [J]. Foreign Language and Culture, 2019(4). 陆强, 罗洪燕, 王丽萍. 大数据背景下的译后编辑项目研究[J]. 中国科技翻译, 2019, 32(3): 35-37, 26. Lu Qiang, LUO Hongyan, Wang Liping. Chinese Science & Technology Translation, 2019, 32(3): 35-37, 26. 王华树，李智．人工智能时代笔译员翻译技术应用调查 —— 现状、发现与建议[J]. 外语电化教学, 2019, 190(6) : 67-72． Wang Huashu, Li Zhi. A Survey on the Application of Translators' Translation Technology in the Era of Artificial Intelligence: Current Situation, Findings and Suggestions [J]. Audio-visual Teaching of Foreign Languages, 2019, 190(6) : 67-72. 吴戈. 机器翻译：语义还是数据？ —— 人工智能视域下的语义问题与哲学反思[J] 河南社会科学, 2019(4): 75-80. Wu Ge. Machine Translation: Semantic or data? -- Semantic problems and philosophical reflections from the perspective of artificial Intelligence [J]. Henan Social Sciences, 2019(4): 75-80. 朱慧芬, 赵锦文, 诸逸飞. 在线机器翻译的译后编辑原则研究 —— 以“八战略”为例[J]. 中国科技翻译, 2020, 33(2): 24-27. Zhu Huifen, Zhao Jinwen, Zhu Yifei. Research on Post-translation Editing Principles in Online Machine Translation -- A Case Study of "Eight Strategies" [J]. China Science & Technology Translation, 2020, 33(2): 24-27. 蔡强, 董冬冬. 基于GOOGLE神经网络汉英翻译的译后编辑研究 —— 以科技文本为例[J]. 西南石油大学学报: 社会科学版, 2020, 22(1): 107-112. CAI Qiang, Dong Dongdong. Research on Post-Translation Editing based on GOOGLE Neural Network for Chinese-English Translation: A Case Study of Sci-Tech Texts [J]. Journal of Southwest Petroleum University: Social Science Edition, 2020, 22(1): 107-112. 熊璨. 论人工智能翻译的可能性——从翻译的三个层次看非文学与文学翻译[J]. 中华文化与文论. 2020, (45): 106. Xiong Can. On the Possibility of ARTIFICIAL Intelligence translation -- Non-literature and Literary Translation from the Three Levels of Translation [J]. Chinese culture and Literary theory. 2020, (45) : 106. 翁义明, 王金平. 文学语篇机器翻译的特征与局限[J]. 当代外语研究, 2020(6). Weng Yi-ming, Wang Jin-ping. Machine Translation of Literary Discourse [J]. Contemporary Foreign Language Studies, 2020(6). 侯桂新.《华盖集续编》校读记 —— 兼谈2005年版《鲁迅全集》第三卷的校勘问题[J]. 中国现代文学研究丛刊, 2020, 04: 85-112. Hou Gui-xin. Proofreading of Huagai Collection Continuation -- Also On The Collation of the Third Volume of The complete Works of Lu Xun (2005) [J]. Modern Chinese Literature Studies Series, 2020, 04: 85-112. 陈伟. 机器翻译对译者主体性的解构 —— 兼论人工翻译的未来落脚点[J]. 外语研究, 2020(2): 76-83. Chen Wei. Deconstruction of Translator's Subjectivity by Machine Translation -- Also on the Future Foothold of Human Translation [J]. Foreign Language Studies, 2020(2): 76-83.] 赵涛. 机器翻译译后编辑的现状与问题[J]. 外语教学, 2021(4). Zhao Tao. Current Situation and Problems of Post-translation editing in Machine Translation [J]. Foreign Language Teaching, 2021(4). 李梅. 机器翻译译后编辑过程中原文对译员影响研究[J]. 外语教学, 2021(4). Li Mei. Research on the Influence of Original Text on Translators in Machine Translation and Post-editing [J]. Foreign Language Teaching, 2021(4). 陈胜, 田传茂. 在线翻译平台汉英翻译的问题及译后编辑 —— 以石油地质文献为例[J]. 中国科技翻译, 2021, 34(1): 31-34, 49. Chen Sheng, TIAN Chuanmao. Problems and Post-translation Editing of Chinese-English translation on online translation platform -- A Case study of petroleum geology Literature [J]. China Science & Technology Translation, 2021, 34(1): 31-34, 49. 郭明阳. 人工智能在机器翻译中的应用研究[J]. 河南科技大学学报（自然科学版）, 2021, 42(3): 97-104. Journal of Henan University of Science and Technology (Natural Science Edition), 2021, 42(3): 97-104. 王湘玲、赖思、贾艳芳. 人工翻译与神经网络机器翻译译后编辑比较研究 ——基于对隐喻翻译的眼动追踪与键盘记录数据[J]. 外语教学理论与实践, 2021(4): 115-126. Wang Xiangling, Lai Si, Jia Yanfang. A Comparative Study on Post-translation Editing between Artificial Translation and Neural Network Machine Translation: Based on Eye Tracking and Keystroke Data for Metaphorical Translation [J]. Foreign Language Teaching Theory and Practice, 2021(4): 115-126.]