Since the 20^th century, with the development of technology and the application of translation technology in translation, machine translation has become a hot spot of attention in translation research today. Today, the demand for translation in various industries has increased, and machine translation has significant advantages in improving translation speed and translation efficiency. However, machine translation still has problems, such as translation quality and translation style, which are difficult to guarantee, so the consequent post-translation editing has also become a hot spot for research and discussion by scholars in the translation field. The combination of machine translation and post-editing can help improve translation quality and translation efficiency. Post-editing (PE) refers to the process of processing and modifying the original output of machine translation according to certain purposes, and it has developed into an emerging profession in the global language service industry [1]. Post-editing is one of the translation processes of machine translation. After the machine translates the original translation, the translator processes and adjusts the machine translated content according to the customer's needs and translation purposes.

Domestic scholars such as Wang Xiangling have compared and analyzed the research data of machine translation at home and abroad, and post-translation editing scholars such as Cui Qiliang [2] have discussed the focus analysis and development trend of post-translation editing research, and all these data analysis have contributed to the development of machine translation and post-translation editing. However, there are few analyses of research hotspots and research trends of domestic post-translation editing applications at present, and understanding the development trend of domestic post-translation editing has important reference value for the future development trend of machine translation, the improvement of post-translation editing ability and the cultivation of translation talents.

Based on this, this paper collates domestic post-editing-related research with the help of CiteSpace, a visualization tool for econometric analysis, with the aim of analyzing the dynamics and research trends of domestic post-editing research, while attempting to provide references for the development of post-editing tools and the cultivation of domestic post-editors.

Research Design

Data sources and Selection Basis: This study used China National Knowledge Infrastructure (hereafter CNKI) database as the literature source, and conducted multiple rounds of literature search with the combined themes of "post-editing" and "machine translation post-editing", and the search time was set from "1995 to 2020". In addition to the non-research literature such as conference announcements and newsletters, as well as the literature that did not include post-editing as the main research object, a total of 205 valid documents were obtained as data samples after manual screening. CiteSpace, a metrological and statistical software developed by Prof. Chaomei Chen, generates visual maps based on literature data, which can visually reflect the development of the field to which the research object belongs and the change of keywords, and it has been widely used in various research fields and produced many results. Therefore, this paper takes CiteSpace as the main metrological tool, supplemented by the built-in statistical function of CNKI, and uses CNKI articles as the literature source to try to visualize and analyze 25 years of post-translation editing research in China, and separately count the publication dynamics, literature sources, high-yield authors, high-frequency co-occurring words, research hotspots and other elements.

Data Analysis Tool

Citespace is an information visualization software developed in Java language, which is mainly based on the co-citation analysis theory and pathFinder algorithm, etc. It can help researchers measure the literature of a specific field and explore the development of the discipline. What’s more, it can find its knowledge inflection point, and through a series of visualization mapping to form. This paper analyzes the research hotspots and research trends of post-editing research at home and abroad by using Citespace 5.7.R5W software, with a view to providing new perspectives for the development of machine translation and the cultivation of translation talents.

The author uploaded the processed literature into CiteSpace software platform according to the operation procedure. The time slicing is from 1995 to 2020. The year per slice is set to 1 year, and the term source selects “Title”, “Abstract”, “Author”, “Keywords”, “Keywords Plus". The node types select “Keyword”, “Author” and “Institution”. And the data analysis threshold is top N = 10, i.e., the top 10 nodes with the highest frequency are selected, and the pathfinder is used to generate the correlation map.

Analysis of the Dynamics of Literature Publication

As Figure 1 shows, the number of machine translation post-editing research publications has shown a gradual and increasing trend. The period from 1995 to 2009 was a stagnant period for the research on machine translation post-editing in China. During this period, only 4 related papers were published, accounting for 2% of the total literature, and the post-editing research in this period tended to explore and think about the machine translation post-editing on a macro perspective. The period from 2010 to 2015 was a slow development period for the research on post-editing. After 2016, the research on post-editing of domestic machine translation marched a period of rapid development, and the overall research trend keeps rising. Moreover, scholars in the translation field pay more and more attention to the research on post-editing. In 2020, the number of relevant papers reached 63, and the research perspectives gradually diversified and focused more on microscopic post-editing strategies, translators' post-editing ability in the era of artificial intelligence, humanistic factors in the process of post-editing of machine translation, post-editing of neural network machine translation, and the cultivation of post-editing talents.

Literature Distribution Analysis

According to Figure 1, among the 205 research papers on machine translation post-editing, 123 of them are master's degree theses, accounting for 60%. And the remaining 82 are from 50 journals. Among them, 12 articles were published in Overseas English, accounting for 5.85%; 6 articles were published in CSSCI core journal Chinese Science & Technology Translators Journal, accounting for 2.93%; 2 articles were published in Chinese Translators, Technology Enhanced Foreign Language Education and English on Campus, respectively, accounting for 1.46% each; 2 articles were published in Contemporary Foreign Language Studies, accounting for 0.98%; and 1 article was published in other 39 journals, respectively. It can be seen that the literature on machine translation post-editing in China mainly comes from master's degree dissertations, and the related literature is mostly found in translation and foreign language professional journals. In addition, 22 articles were published in core journals or CSSCI journals, accounting for 10.7%, among which there are many authoritative journals in Chinese translation industry, such as Chinese Translators, Shanghai Journal of Translators, Chinese Science & Technology Translators Journal, and Foreign Language World. However, in general, the research on post-editing still lacks high-level and high-quality research results. Thus, there is an urgent need for researchers to conduct post-editing research in a broader, more diversified and deeper way.

High Influential Author Analysis

High influential authors are a group of authors with certain academic influence in a certain discipline, and they often represent the orientation and trend of discipline development. By studying the current high-influential authors in the field of post-editing research in China and understanding their scientific research practices, it is easy to grasp the development trends of post-editing research and promote the development of machine translation post-editing in China. 

Figure 1: The number of post-editing publication from 1995 to 2020

Table 1. The Source of Post-editing Publication from 1995 to 2020（Partial）

Figure 2: Collaboration among High Productive Authors on Post-editing (1995-2020)

The number of publications and citation frequency can reflect the research productivity and academic influence of authors. In this paper, we will analyze the high influential authors in the field of post-editing in China in terms of the number of publications.

Since most of the authors published fewer articles, and the number of articles published by the authors from the same institution is small, in order to highlight the few researchers who concentrated on post-editing and related research institutions.

Table 2: The Author of Post-editing in China (1995-2020)

The threshold value was limited to be greater than or equal to 2, and the relevant researcher map was derived (Figure 2). The knowledge map in Fig. 2 includes 134 nodes, 32 connecting lines, and a density of 0.0036. In this map, the font size varies as a whole, and the person corresponding to the largest font means that he or she has published the most articles, where there are collaborative relationships among researchers. Then there are connecting lines in the middle, indicating the collaborative relationships between them. The larger nodes of Feng Quanguong, Cui Qiliang, Zhang Huiyu, Shi Wenjie and Zhou Wenge represent the greater influence of these authors in domestic post-editing research, and they have promoted the development of machine translation application research in the domestic translation industry. Among them, Feng Quankong has published 6 articles and Cui Qiliang has published 3 articles, revealing that these two scholars have been committed to the research on post-editing. The existence of linkages between Feng Quanguong, Zhang Huiyu and Cui Qiliang, and Zhang Chunbai and Wei Changhong scholars indicates that these scholars have been cooperating more closely in the research of post-editing. However, the nodes are more scattered and less connected, indicating that the collaborative research among scholars is not close enough and the interdisciplinary collaborative research can be further strengthened.

We selected eight scholars with the number of publications (≥2) and counted their first publication year, i.e., the year when they first published papers related to machine translation application research (see Table 2). Through sorting, the related research in China can be summarized as follows: systematic research on post-translation editing and post-translation editing and language services.

Post-editing System Research

Before 2015, scholars in China mainly focused on the concept of machine translation and the improvement of its system. Huang Heyan and Chen Zhaoxiong [3] were the first to propose the design principle and implementation algorithm of an intelligent post-editing in 1995, which takes the meaning segment as the basic processing unit. And it not only can form post-editing feedback information suitable for reverse reasoning but also provide processing basis for self-improvement of machine translation system knowledge. Morever, it can contribute to improve the efficiency of post-editing by setting automatic adjustment mechanism of multiple meaning segment translation positions. Wei Changhong and Zhang Chunbai [4], as one of the early explorers of post-editing of machine translation in translation science, put forward the necessity of post-editing in the article "Machine Translation Post-editing" co-authored by them in 2007, while manual post-editing is an essential part of machine translation [4]. In order to clarify the concept of post-editing, Cui Qiliang, in his article "On Post-editing of Machine Translation" published in 2014 [2], systematically expounded the concept of post-editing, the relationship between machine translation and post-editing and the application of post-editing, which pointed out the way for the subsequent research and development of post-editing [4]. With the acceleration of globalization, especially after China proposed “One Belt, One Road”, the international and domestic foreign exchanges and cooperation have strengthened. In addition, the demand for language services from customers has surged, and translation and translation technology have been combined. Post-editing has emerged, and it will play an increasingly important role in the language service industry.

Post-editing and Language Service

Feng Quanguong and Zhang Huiyu [5] were one of the first to propose the introduction of post-editing into translation talent training and translation education in China, and explored the cultivation of post-editors from the perspective of post-translation editor cultivation research in the context of global language service industry. They discussed the industry demand for post-editors, post-editing ability, post-editing curriculum, post-editing teaching and post-editing tool selection. However, in recent years, with the rapid development of machine translation and translation technology, post-editing has become more and more important, and the related application research on post-editing has only gradually attracted the attention of the translation academia. In the article "Post-editing research: focus analysis and development trend" co-authored by Feng Quankong and Cui Qiliang in 2016 [1], the development trend of post-editing research was analyzed, which includes the development of integrated translation work environment, the development and application of specific machine translation system, the cultivation of post-editing talents, and the in-depth cooperation between industry, academia and research, etc. The domestic demand for language services is increasing. The application of machine translation has greatly improved the translation efficiency, but the translation quality is difficult to guarantee. Therefore, the combination of post-editing and machine translation has become the current trend in the language service industry. Wang Xiangling and Jia Yanfang [6], on the other hand, analyzed the research progress and research methods by comprehensively analyzing four research hotspots: post-editing process and product evaluation, post-editing efficiency influencing factors, post-editing tools and post-editors and personnel training. It expounds its future development trend and its revelation to the research of post-editing of machine translation in China, so as to provide a new perspective and method for domestic related research as well as translation personnel training. This paper will provide new perspectives and methods for domestic research, translation talent training and construction of translation disciplines. With the rise of post-editing, post-editing competence has become an emerging topic in the field of translation research. Feng Quankong and Liu Ming [7] proposed the construction of a three-dimensional model of post-editing ability containing cognitive dimension, knowledge dimension and skill dimension, and elaborated the specific composition and representation of each dimension. Zhong Mingming and Shu Chao [8] analyzed the existing post-editing teaching samples in foreign countries and constructed a teaching-oriented post-editing competency structure from the actual needs of post-editing teaching, which provided a new perspective for the domestic post-editing curriculum. Although there is not much relevant literature on post-editing research in China, from systematic research on post-translation editing to the current exploration of combining post-editing with translation talent training, the research perspectives and research methods have shown diversification, laying a foundation for post-editing research, post-editing competence and post-editing strategies, as well as providing a training direction for the cultivation of future university translation talents.

Keyword Co-occurrence Analysis

Keywords not only reflect the research topic but also highly summarize the research content. Analyzing the keywords of the paper can understand the research hotspots in the field. Each node in the keyword knowledge graph represents a keyword, and the size of the node is proportional to the frequency of the keyword. The more frequently a keyword appears, the larger the node is, and vice versa. The lines between keyword nodes indicate the co-occurrence relationship between keywords. And the thicker the lines are, the higher the frequency of keyword co-occurrence. In this paper, the author uses Citespace V5.7.R4W information visualization software to count the frequency of keywords. It sets Node Types to Keywords, and the Node Shape is "Circle" to represent keywords. Its Threshold is greater than or equal to 4. The node size in the graph is determined by the frequency of keywords in the literature, and the larger the node size is, the more frequently the keyword appears in the literature. The font size of the keyword then indicates the strength of its centrality, i.e., the role the node plays in connecting other nodes, and the stronger the centrality indicates the more research conducted through the keyword, the greater its influence in the co-occurrence network. The line between keywords represents co-occurrence relationship, and the thicker the line is, the higher the frequency of co-occurrence of keywords and the closer the connection.

Research Hotspots Analysis

Figure 3 shows the co-occurrence network knowledge map based on high-frequency keywords in China during 1995-2020, in which the network nodes are 302, the number of connected lines is 887, and the density is 0.0195. Keywords with a frequency greater than 4 are selected in this paper, and the top 14 ranked keywords are listed in detail in Table 3. These keywords represent the post-editing research themes that domestic research scholars focused on during the period 1995-2020. From Table 3, it can be seen that the earliest research related to post-editing in China appeared in 1995. It indicates that with the development of machine translation and translation technology in China, the combination of translation and translation technology has provided a new research perspective for translation research. And scholars have started to actively devote to the research related to post-editing of machine translation. Although machine translation can improve translation efficiency and effectively shorten translation time, it still cannot avoid the problem of translation errors, which requires post-editing to improve translation quality and correct errors. From the earliest systematic research focusing on post-editing, domestic scholars are now paying more and more attention to the research on post-editing strategies and the cultivation of post-editing ability.

Figure 3: Keyword co-occurrence network map of post-editing research in China between (1995-2020)

Figure 4: Machine translation post-editing burst keywords

Table 3: 1995-2020 Keywords and its centrality of Post-editing Studies in China

Frontier Analysis

Burst Keywords refer to keywords that appear suddenly or proliferate within a certain period of time, and can detect dynamic changes and developmental frontiers in related research fields. In this paper, we use the Burstness/View control panel function of CiteSpace software to derive 12 burst keywords for post-editing research from 1995 to 2020 (Figure 4). Translation quality, pre-editing, language service industry, post-editors and computer-assisted translation were once the frontiers of research in this field. Translation quality research was the emergent keyword in 1995-2014, pre-editing emerged from 2005 to 2014, language service industry research emerged in from 2015 to 2019, and post-editors emerged from 2015 to 2019. After 2015, "English-Chinese translation", "translation quality" and "translator's attitude" have higher emergence intensity values with emergence periods of 2015-2020, 2015-2020, 2015-2020 and 2015-2020, respectively, is the frontier of post-editing research for machine translation in recent years.

Limitations and Prospects

With the increasing openness of China to the outside world and the more frequent international economic and trade cooperation, the translation quality and translation requirements of the language service industry have become higher and higher. In recent years, with the research and development of translation tools, machine translation has been the focus of research in the translation field, and has been paid attention to by a wide range of scholars and experts, and the post-translation editing that comes along with machine translation has received great attention, and the number of published core journal papers has shown a trend of growth with each year, but the analysis of Citespace visual data reveals that the research on post-translation editing of machine translation currently has the following shortcomings:

Single Research Model

According to the statistical analysis above, most of the studies investigates post-editing from the macro level are systematic introductions to the concept of post-editing, while it lacks of studying post-editing from the micro level, with a single research perspective. Many studies on post-editing have been carried out in China, and the topics focus on the overview of post-editing, post-editing evaluation, machine translation error identification, and post-editing tool development [1]. However, these studies are still at the stage of macro-theoretical research, lacking more in-depth post-editing-related practical research and failing to provide specific operational standards for post-editors.

Inadequate Post-Editing Evaluation Studies

Post-editing evaluation studies refer to studies that assess the quality, efficiency, value, meaning, time spent, and cognitive load of post-editing. And the empirical and comparative models are the most common [1]. Through the above data analysis, it is found that most of the existing post-editing studies focus on the analysis of machine translation post-editing practices, which is more ponderous and fragmented and lacks systematicity. Those studies are related to the evaluation of commercial benefits brought by post-editing, the measurement index of cognitive effort spent on post-editing, the efficiency of post-editing, the measurement of various post-editing tools, the factors affecting the efficiency of post-editing in a collaborative translation environment, the relationship between the characteristics of the original text and the time and technical effort spent on post-editing, the influence of translators' professional experience on the efficiency of post-editing, and the relationship between pauses and cognitive effort in post-editing [1]. Theoretical studies on post-editing evaluation in China have yet to be systematically constructed. 

Lack of Post-Editing Skills and Post-Editor Development

Post-editing ability refers to the knowledge system and cognitive literacy required to process and modify the original output of machine translation, according to certain purposes and requirements. [11] The language service industry has started to provide post-editing training services (mostly online training), but few universities in China and abroad are offering post-editing courses. That is mainly due to the lack of teachers or the lack of understanding of post-editing (especially in China). A few universities, such as Zhejiang University, have this course in their training programs for translation majors, but the related teaching work has not been really implemented [7]. Post-editing courses are less offered in China, so the relevant curriculum is more discussed from the perspective of doctrinal thinking. And there are relatively few studies based on the practical experience of teaching post-editing. In fact, foreign countries, especially European regions, have accumulated considerable experience in the teaching practice of post-editing [8]. At present, the cultivation of post-editors is still at the stage of teaching mode exploration. Few MTI majors in colleges and universities offer post-editing courses, and teachers are scarce. There is a lack of empirical research on post-editing course settings. Translation and translation technology are not effectively combined. This is not conducive to the cultivation of translation talents and cannot meet the market demand of language service.

Machine translation error recognition and post-editing tools are not sufficiently developed. Current general-purpose machine translation systems, such as Google Translate, Baidu Translate, Microsoft Bing, etc., can only handle plain text (TXT) files, which brings great inconvenience to post-editing of machine translation [1]. Machine translation error recognition can be a specialized post-editing tool. However, at present, there is a lack of dialogue and communication between translators and experts in the field of translation technology. And the two sides of research results rarely learn from each other, which inevitably causes a disconnection between the research and development end of technology and the use end of tools [7]. At present, there are not many post-editing tools in the language service market that can assist translators in post-editing. Moreover, the more frequently used post-editing tools are SDL Trados and Yicat, etc. However, the functions of these tools are not perfect and need to be further optimized. They still cannot meet the current needs of translators. Based on the above shortcomings, this paper proposes that post-editing can be studied more deeply from the following points: 

Post-Editing Evaluation System Construction

It is important to explore the post-editing assessment system, which will provide assessment criteria and practice guidelines for post-editors and post-editing quality. Cui Qiliang [2] studied the practical guidelines of post-editing in terms of translation quality requirements, types of machine translation systems, identifying error characteristics of machine translation output and the purpose of post-editing. Lu Qiang et al. [9] study how to develop universal post-editing quality specifications based on the universal claims of translation accuracy, completeness, and terminology consistency in terms of the realistic requirements of machine translation and post-editing projects. Based on Newmark’s translation theory, Zhu Huifen et al. [10] review the types of mistranslation in machine translation and explore the post-editing principles of translators at the textual, referential, articulation, and natural levels. Establishing a complete post-editing assessment system not only provides translators with a guide to post-editing actions, but also facilitates the cultivation and practice of new types of translation talents as post-editors.

Post-Editing Education

Wei Changhong and Zhang Chunbai [4] suggested that post-editing is usually practiced by translators. Post-editing relies heavily on bilingual skills accumulated over time, but translators have an advantage over others in this regard. Feng Quankong and Zhang Huiyu [5] discussed the cultivation of post-editors in terms of the industry demand for post-editors, post-editing curriculum, post-editing teaching and post-editing tools selection, and called for the establishment of a separate post-editing course in universities or the integration of post-editing modules into computer-assisted translation courses. At present, the most important task of post-editing teaching is to let undergraduate translation majors and MTI students understand post-editing, know the market of the industry, and have some post-editing practice. Feng Quankong and Liu Ming [7] construct a three-dimensional model of post-editing ability from three dimensions: cognition, knowledge and skill, so as to explore the teaching of post-editing and provide references and instructions. It is a useful attempt to implant the core content of post-editing as an important module in the courses related to computer-assisted translation and localization. And then it offers a separate course on post-editing when the conditions are mature. Zhong Mingming and Shu Chao [8] analyzed the post-editing curriculum in foreign universities on the basis of the frontier analysis of post-editing courses abroad, and constructed a teaching-oriented post-editing competence structure from the actual needs of post-editing teaching, which provides a perspective for the domestic post-editing curriculum. Research on post-editing competencies and teaching is still relatively scarce and needs to be explored in depth.

A post-editing tool development model in which translation technicians and translators cooperate. Li Mei and Zhu Ximing [11] analyzed the output of machine translation with reference translations, machine translation errors, error types and typical lexical and syntactic errors. They also described them formally in the framework of generative grammar, which could then be used to develop post-editing software later. Cui Qiliang [2] proposed that the post-editing environment provided by computer-aided translation systems is fixed, but different users have different needs for post-editing assistance tools, assistance modes and interface styles, etc. The collaborative translation system should be able to predict different users' different needs for post-editing assistance modes at different translation stages according to their interaction behaviors and translation states. It provides the system's assistance in a timely manner to users. The model of combining machine translation and post-editing is rapidly developing, which gives full play to both the speed (efficiency) of machine translation and the accuracy (quality) of human translation. This model not only meets the rapidly developing needs of the translation market, but also promotes the development of translation technology and the communication and cooperation between academia and industry. And it enriches the composition of the language service industry chain. Feng Quankong and Liu Ming [7] believe that translators need to master special knowledge, such as knowledge related to post-editing, the industry demand for post-editing, the development and application of post-editing tools, the difference and connection between post-editing and translation revision, and the feedback requirements of post-editing for improving the technical level of machine translation.

In the context of the era of big data, with the deepening of globalization and the increasing demand for language services, it becomes especially important to improve the quality of machine translation and translation efficiency. And post-editing plays an increasingly important role in machine translation. The translation mode combining machine translation and post-editing has become the current translation mainstream. The training of post-editing ability and post-editors and the research and development of post-editing tools should be strengthened. In addition, the post-editing evaluation system and practice guidelines should be formulated.

1. Feng, Quangong and Qiliang Cui. "Post-editing studies: Focused analysis and development trends." Shanghai Translation, no. 6, 2016.

2. Cui, Qiliang. "On post-editing in machine translation." Chinese Translators Journal, no. 6, 2014.

3. Huang, Heyan, and Zhaoxiong Chen. "Design and implementation algorithm of an intelligent post-editor." Journal of Software, no. 3, 1995.

4. Wei, Changhong, and Chunbai Zhang. "Post-editing in machine translation." Chinese Science & Technology Translators Journal, no. 3, 2007.

5. Feng, Quangong, and Huiyu Zhang. "Research on the training of post-editors in the global language services industry." Foreign Language World, no. 1, 2015.

6. Wang, Xiangling, and Yanfang Jia. "Empirical studies on post-editing in MT abroad in the 21st century." Journal of Hunan University (Social Sciences), vol. 2018, no. 2, pp. 82–87.

7. Feng, Quangong, and Ming Liu. "Construction of a three-dimensional model of post-editing competence." Foreign Language World, no. 3, 2018, pp. 55–61.

8. Zhong, Wenming, and Chao Shu. "Competence structure and curriculum design for post-editing—A frontier analysis of foreign courses." Computer-Assisted Foreign Language Education, no. 6, 2020.

9. Lu, Qiang et al. "Research on post-editing projects under big data context." Chinese Science & Technology Translators Journal, vol. 32, no. 3, 2019, pp. 35–37.

10. Zhu, Huifen et al. "Principles of post-editing in online machine translation." Chinese Science & Technology Translators Journal, vol. 33, no. 2, 2020.

11. Li, Mei and Ximing Zhu. "Exploring automation of English-Chinese MT post-editing." Chinese Translators Journal, no. 4, 2013.