The "machine translation + post-machine translation editing" model on the YiCAT platform: using the example of the discipline "Stylistic post-machine text editing" for first-year graduate students of the Higher School of Translation of Moscow State University

Lyu Jingpeng PhD in Philology Lecturer; Higher School of Translation; Lomonosov Moscow State University 108802, Russia, Moscow, Krasulinskaya str., 21, sq. 104 ljpesti@mail.ru

Introduction

Machine Translation (MT), also known as automatic translation or computer translation, is the process of using computers to convert text in one natural language (source language) into text in another natural language (target language). With the rapid development of Internet technologies, as well as the rapid expansion of the amount of information presented in various languages, machine translation technology gradually ceased to be exclusively entertaining, becoming an effective tool for overcoming natural language barriers in society. In the context of the gradual development of machine translation technology, more and more researchers are paying attention to the question of the possibilities of post-machine translation editing ^[1].

Nowadays, machine translation in many cases helps or even replaces a person when translating. However, despite the rapid development of this technology, machine translation still has certain qualitative disadvantages. In practice, it is difficult for translators to perform a large amount of translation work in a limited time in modern working conditions, using only human resources. Instead, many translators use machine translation tools, as well as perform post-machine translation editing. Thus, post-machine text editing is a necessary component in the work of modern translators. However, both in Russia and in China there are gaps in the field of building systems for assessing the complexity of post-machine editing, designing auxiliary translation systems, as well as developing and researching post-machine editing tools. In addition, there are only a few hundred articles in Russia that address the specifics of machine translation. Most of them are devoted to the study of existing limitations, development opportunities and ways to improve machine translation systems. At the same time, there are extremely few articles dealing with the features of post-machine editing of the translation text. The number of such articles does not exceed several dozen. Thus, there are many challenges in the field of post-machine translation editing research that should be addressed as soon as possible so that the field of post-machine translation editing can correspond to the era of artificial intelligence. At the same time, Russian universities should pay attention to the teaching of disciplines affecting machine translation tools and post-machine editing skills. Large translation faculties should actively use translation technologies and machine translation tools in the classroom, add new academic disciplines implying the familiarization and application of such technologies and tools by students, as well as integrate such disciplines into the framework of existing training programs for specialists in the field of translation. All this will allow universities to form a reliable basis for the education of modern and qualified specialists who meet the requirements of the new era. This issue has already been updated some time ago. So, in the fourth issue of the journal "Bulletin of Moscow State University" for 2019, the dean of the Faculty of Higher School of Translation, N. K. Garbovsky, together with the deputy dean of the Faculty, O. I. Kostikova, published an article "INTELLIGENCE FOR TRANSLATION: ARTFUL OR ARTIFICIAL?" The authors of the article state the following: "Today, a new concept is being born — "digital translation", which defines a new type of translation technology, a system of network interaction between a translator and digital information and communication tools, artificial intelligence (AI), designed to increase the effectiveness of translation art and the quality of translation products. Translation into the digital age is a complex system of contradictory relations in the binomial "man — smart machine" ^[2]. In addition, in 2022, N. K. Garbovsky and O. I. Kostikova formed an educational and research group. The group included 3 teachers of Chinese at the Moscow State University Graduate School of Translation and 4 graduate students. As a result of the research, a new discipline for Chinese first-year master's degree students called "Stylistic post-machine text Editing" was launched on February 8, 2022. The reason for the creation of this discipline is to meet the development trends of the era of artificial intelligence, as well as to update the training program for specialists to develop comprehensive translation competencies that meet the needs of modern digital technologies.

1.Organization of the discipline "Stylistic post-machine text editing"

In preparing the discipline program, the authors studied and analyzed many sources in Russian and Chinese, including scientific literature and articles that touched on the process of learning the skills of post-machine translation editing. It was found that most of the existing research focuses on errors and strategies of English-Russian machine translation. A small amount of research has been devoted to translation errors and strategies within the framework of Chinese-Russian machine translation. There are only a few studies on the process of learning to edit Chinese-Russian translation post-machine, as well as the types of errors that occur in the process of post-machine editing, and methods for correcting them. At the same time, the presented studies mainly used machine translation tools such as Google Translate and Deepl Translator, as a result of which these studies have a single object, while the results obtained are not widely used. Since different languages have different linguistic characteristics, and different translation systems have different translation characteristics, the types of errors in machine translation differ.

The program of education and professional training in this discipline should cover such aspects as machine translation technologies, post-machine translation editing strategies, linguistic knowledge, etc. There are some studies that consider students' attitude to post-machine translation editing and their user experience, as well as ways to develop effective post-machine editing training programs ^[3]. These studies have become a source of important information about the possibilities of post-machine editing in the preparation of the relevant discipline.

"Stylistic post—machine text editing" is a mandatory second semester course for first-year master's degree students, which gives 2 credits and implies mastering the program within 72 academic hours (32 academic hours are allocated to lecture-type and seminar-type classes; 40 academic hours are allocated to independent work of students).

The content of the discipline is divided into the following modules:

1. Familiarization with existing machine translation programs, comparative analysis of the advantages and disadvantages of each program;

2. Post-machine editing of the translation on the topic "news";

3. Post-machine editing of the translation on the topic "public texts";

4. Post-machine editing of the translation on the topic "cooking and restaurant menu";

5. Post-machine editing of the translation on the topic "instructions";

6. Post-machine editing of the translation on the topic "description of the city";

7. Post-machine editing of the translation on the topic "business correspondence".

The results of training in the discipline:

The final exam in the discipline is conducted in the form of a credit with the appropriate assessment options: credited / not credited.

2. Educational platform for the discipline "Stylistic post-machine text editing"

YiCAT (official YiCat website: https://www.yicat .VIP) is an online translation management platform developed by Shanghai Yizhe Information Technology Co., Ltd. (Tmxmal), which aims to provide users with faster and more efficient translation and localization solutions. Since its launch, the YiCAT platform has maintained a professional level of service, and new solutions are being developed in areas such as multilingual big data analysis, machine translation using artificial intelligence, natural language processing, intelligent text processing and file processing in various formats, etc. The YiCAT platform offers users a variety of products to choose from. including the group version, the corporate version and the university version of the platform. Any version of the platform can be used for free on request, which is very convenient for users. As a result of the comparison of several similar platforms, we came to the conclusion that YiCAT is indeed a suitable tool for teaching and practicing within the discipline of post-machine editing. This platform is suitable for teaching translation based on real translation projects, which allows teachers to perform the task of teaching in the most qualitative and systematic way, as well as properly educate specialists combining both linguistic and necessary technical skills. The key reasons for choosing this platform are as follows:

1. Translation practice platform, ideal for students

The translation process is displayed visually. Teachers can check the progress of students' assignments at any time. They can set up various options, such as self-assessment of students, mutual assessment and teacher assessment, as well as accumulate a translation corpus during the practice process.

2. Integration of multiple machine translation systems

The platform is connected to dozens of Chinese and foreign machine translation systems, such as Google and Baidu, between which there is a smooth switching during operation. The platform supports the process of editing machine translation, as well as the function of quickly calling machine translation tools to meet the needs of individual projects and tasks.

3. Deep integration of artificial intelligence technologies

In combination with artificial intelligence algorithms, the platform can automatically fill the translation with special marks, automatically optimize the results of machine translation, as well as integrate the specified terminological database or memory database into the results of machine translation, allowing the user to more conveniently implement translation into work.

4. Support for 30 major file formats

The platform supports 30 file formats such as doc/docx, xls/xlsx, ppt/pptx, pdf, txt, html, etc.

5. Availability of 46 languages

The platform uses 46 languages, including: Chinese, English, Japanese, Korean, Russian, German, Spanish and Portuguese.

6． Saving the file format

The platform supports exporting to 30 file formats. The exported design and content remains unchanged, while opening and reading documents becomes more user-friendly.

7. Saving translation records

The platform keeps records of the transfers made. No additional payment is required for re-downloading.

8. Converting PDF documents

The platform supports the function of converting PDF to Word, PPT and Excel formats with accurate preservation of the content and design of the source file.

9. Simplification of the learning process and integrated management

With the help of this platform, the teacher can combine the development of training courses and programs, form classes, prepare homework and exams, make translation adjustments, and use other functions. All this allows you to make the translation learning process easier and implement it in a more closed format. The management process looks like this:

The maximum number of students is more than 500 people; the number of files in one project is 300; the number of proposals in one memory database is 3,000,000; the number of entries in the terminology corpus is up to 1000.

Stages of training in the framework of the discipline "Stylistic post-machine text editing"

1. Effective allocation of study hours and the division of 16 classes into five stages of training.

The introductory stage (1st and 2nd weeks of training). Students get acquainted with the process of the YiCAT platform and machine translation tools, perform simple exercises on machine translation of news articles, and also edit machine translation. At this stage, the task is to give students a faster way to get used to using the platform. It is assumed that students have previously met with news articles during their undergraduate studies. Regardless of whether students practice listening or reading, most choose the news style that is required when translating news texts.

Since students are only at the introductory stage of the discipline, in order to reduce the excessive anxiety associated with learning, the teacher can reduce the complexity of tasks and give tasks for group performance. It will also allow students to develop teamwork skills.

The initial stage (from the 3rd to the 6th week of training). Students are already familiar with the rules of using the platform. Students may try to translate basic texts such as public texts or restaurant menus. Texts of this type are closely related to the life of students and are quite simple in terms of grammatical structure, which makes them moderately difficult to translate. This approach allows students to increase their interest in editing machine translation.

Advanced stage (from the 7th to the 10th week of training). In order to implement a smooth learning process and gradually increase the level of complexity, the content of translation tasks is changing at this stage. In particular, there is a transition from the translation of short public texts to longer sentences presented in the instructions. When translating instructions, machine translation has many significant advantages. For example, with the help of linguistic corpora or a database of big data, the machine can accurately select the desired translation option for individual terms, which significantly increases the efficiency of the translator.

The final stage (from the 11th to the 15th week of training). After 10 weeks of study, students can switch to the practice of translating texts of various styles. For example, a translation on the topic of urban description or business correspondence is possible. Teachers check the final translation, after which they leave their comments. In addition, teachers make translation adjustments, make marks and choose the best translation using special tools on the platform. During the discussion of translations in the classroom, all students share their translation. Students summarize the types of machine translation errors and suggest appropriate translation strategies for each incorrect variant. Based on certain translation strategies, students re-view the translations of each group and compile reports, pointing out errors that occurred after their own editing of machine translation.

During the control period (week 16), oral and written exams are organized at the end of the semester.

Conclusion

Translation tools based on the use of artificial intelligence are a modern and fast translation method that increases the efficiency of the translator's work. However, this method has a number of definite disadvantages. Currently, post-machine translation editing is especially important. At the same time, systematic training in translation strategies and fostering more developed thinking among translators performing post-machine translation editing is an even more important task^[4].

With the help of the YiCAT platform, students can take courses in post-machine translation editing, which significantly increases the effectiveness of their training and allows teachers to track the progress of students completing assignments in real time. In addition, data archiving and communication of participants in the learning process takes place on the platform, which allows teachers to evaluate students' results and academic performance at any time. Data on the learning process also allows us to trace patterns in students' acquisition of knowledge about translation, which allows teachers to adjust educational materials taking into account students' mistakes in translation. This significantly optimizes the effectiveness of training. Compared to traditional teacher-led learning models, students feel more satisfied when using the online learning model of post-machine editing, which also increases their sense of belonging. Thus, the model, which uses a combination of machine translation and post-machine editing, lays the foundation for the development of translation specialists required in a new era.