|
Library
|
Your profile |
This work is licensed under a Creative Commons Attribution-NonCommercial 4.0 International License.|
Litera
Reference:
Liu, M., Shao, Q., Xie, G. (2024). Multi-Agent Approach to Political Discourse Translation: From Large Language Models to MAGIC-PTF System. Litera, 11, 28–46. https://doi.org/10.25136/2409-8698.2024.11.72197
Multi-Agent Approach to Political Discourse Translation: From Large Language Models to MAGIC-PTF System
DOI: 10.25136/2409-8698.2024.11.72197EDN: GFRZMOReceived: 01-11-2024Published: 08-11-2024Abstract: This research addresses the automated translation of Chinese political discourse into Russian utilizing Large Language Model (LLM) optimization technologies and a multi-agent approach. The study focuses on developing the MAGIC-PTF system, which implements multi-stage text processing through the interaction of four specialized agents. The system's key component is the Style Agent, which ensures stylistic uniformity and terminological accuracy based on a specifically trained LLM. The Translator Agent performs the primary translation work and is responsible for the final text formatting. The Editor Agent conducts multi-level verification and correction of translations, considering linguistic, semantic, and cultural aspects. The Reader Agent analyzes the text from the target audience's perspective, evaluating its reception by native Russian speakers. The methodology integrates LLM optimization technologies and a multi-agent approach, with experimental testing conducted on the fourth volume of "Xi Jinping: The Governance of China" and its official Russian translation. The study includes a comprehensive analysis of system effectiveness using the COMET metric and comparative testing with existing machine translation platforms. The research's scientific novelty lies in developing a methodology for applying LLMs to specialized translation tasks and creating an effective coordination mechanism for intelligent agents in the translation process. Experimental research demonstrated MAGIC-PTF's superiority over traditional machine translation systems in key parameters: terminological accuracy, stylistic consistency, and preservation of culture-specific elements in political discourse. The developed system opens new possibilities for automated translation of political discourse and can be adapted for translating other specialized text types, confirming its significance for modern translation technology development. Of particular value is the system's scalability and adaptability to various language pairs and discourse types, creating prospects for further development in automated specialized text translation. The research findings also contribute to advancing the theory and practice of LLM application in professional translation. Keywords: Large Language Models, Multi-Agent Approach, Political Discourse, Chinese-Russian Translation, LLM Optimization Technologies, Intelligent Agents, Automated Translation, Specialized Translation, Cross-cultural Communication, Machine TranslationThis article is automatically translated. Introduction Political discourse is an essential tool for deepening mutual understanding between States and peoples. In the context of relations between the Russian Federation and the People's Republic of China, which, according to the joint statement of the two countries, have entered a new era of comprehensive partnership and strategic interaction, accurate and adequate translation of political texts is of particular importance. It is through the prism of political discourse that an understanding of strategic goals, value orientations and development prospects of partner states is formed. In the context of the intensification of Russian-Chinese cooperation in all spheres, traditional approaches to the translation of political texts face certain limitations both in terms of ensuring the necessary translation volumes and in terms of the quality of the transfer of semantic nuances. This necessitates the development and implementation of innovative technological solutions that can improve the efficiency and accuracy of translating political discourse from Chinese into Russian. In the context of the rapid development of artificial intelligence technologies, special attention is drawn to large language models (BYAM), the appearance of which [1];[2] marks a qualitatively new stage in the development of computational linguistics and translation technologies. By demonstrating impressive results in various tasks of natural language processing [3], BYAM opens up fundamentally new opportunities for automation and optimization of translation activities. Nevertheless, in such a specific area as the translation of Chinese political discourse into Russian, the results of BYAM's work have not yet reached the level of official translations. The main discrepancies are observed in three key aspects: stylistic correspondence to the genre of the political text, the accuracy of the transfer of special terminology and an adequate interpretation of the deep semantic structures characteristic of political discourse. These limitations necessitate the development of specialized approaches to the use of BYAM in the field of political translation. In the modern scientific literature, research on the use of BYAM in translation activities is developing in two main directions. The first direction, developed from the standpoint of linguistics, is characterized mainly by a review-analytical approach to assessing the potential of BYAM and the prospects for their application [4-6], as well as exploring the possibilities of integrating BYAM into training programs for professional translators [7-9]. The second direction, represented by the works of specialists in the field of computer science, focuses on the technological improvement of the translation process [10-13]. However, the growing complexity and specialization of translation tasks show that neither a purely linguistic nor an exclusively technological approach is capable of fully realizing the potential of artificial intelligence technologies in the field of translation. In the light of the above, the development of an integrated approach combining the linguistic foundations of translation with the latest methods of optimization of BYAM is of particular relevance [14];[15]. As part of this integration, we propose a multi-agent interactive translation system MAGIC-PTF (Multi-Agent Interactive Chinese Political Translation Framework), designed to improve the quality and effectiveness of translating Chinese political discourse into Russian using BYAM. The developed MAGIC-PTF system, operating on the basis of ChatGPT [16], includes four specialized agents: a Stylist Agent, a Translator Agent, an Editor Agent and a Reader Agent. Their interaction is carried out according to a specially developed algorithm that provides step-by-step optimization of translation in order to achieve maximum accuracy, naturalness and pragmatic adequacy of the Russian-language text. 1. Current problems of translating Chinese political discourse into Russian In recent years, there has been a significant expansion of Russian-Chinese cooperation, as well as ties between China and the countries of Central Asia. This cooperation has gone beyond traditional diplomatic relations and has covered almost all spheres of public life — from central to local levels of government. As a result, the volume of political texts requiring high-quality and prompt translation into Russian has increased significantly. 1.1. The practice of translation before the advent of large language models Before the widespread introduction of large language models into translation practice, there were two main approaches to translating Chinese political discourse into Russian: traditional translation performed by professional translators and machine translation followed by post-editing. The first approach was characterized by low efficiency and did not allow processing large volumes of text in a short time. The second approach, although it provided higher productivity, did not guarantee proper translation quality and stylistic compliance, while the editing process often led to an avalanche of edits. The fact that translating political discourse requires not only fluency in Chinese and Russian, but also a deep understanding of the political system and cultural context of both countries is particularly difficult. The duration and high cost of training such specialists have led to an acute shortage of qualified translation staff in this field. 1.2. The practice of translation before the advent of large language models Since 2022, the introduction of large language models has significantly increased the efficiency of translation from Chinese into Russian. Extensive training data and improvements in the architecture of neural networks have allowed models to generate translations of sufficiently high quality in an interactive mode. However, the specifics of political discourse have revealed a number of significant limitations of these models. Firstly, the models do not fully reproduce the official business style of political texts, creating translations that are either stylistically neutral or conversational. Secondly, there are deviations from the established terminological equivalents typical of Chinese political discourse. Thirdly, when working with voluminous texts of a complex logical structure, the models demonstrate a lack of understanding of the context and a violation of the coherence of the presentation. 2. MAGIC-PTF: solving urgent translation problems As part of this research, the interactive multi-agent translation framework MAGIC-PTF (Multi—Agent Interactive Chinese Political Translation Framework, hereinafter referred to as MAGIC-PTF) has been developed, aimed at solving the existing problems of translating Chinese political discourse into Russian. The framework is based on the synthesis of BAM optimization technologies and standardized requirements for the translation of political discourse. This approach not only maximizes the potential of BYAM in text generation, but also ensures strict compliance of translations with established norms and procedures, which together contributes to improving both the efficiency and quality of translation of political texts. 2.1 General overview of MAGIC-PTF The functioning of the MAGIC-PTF system is based on the interaction of four specialized Agents developed on the basis of BYAM. It should be noted that the concept of "Agent" is important in the theory of artificial intelligence and is defined as a computing unit capable of independently analyzing the environment, performing logical operations and making purposeful decisions [15];[17]. The developed system includes four interrelated components: a Stylist Agent responsible for the stylistic design of the text, a Translator Agent who performs the main translation work, an Editor Agent who checks and corrects the translation, and a Reader Agent who evaluates the text from the perspective of the target audience. The structure of the system and the mechanism of interaction between Agents are schematically presented in Figure 1.
Figure 1. MAGIC-PTF architecture and Workflow 2.2 Theoretical foundations of MAGIC-PTF development 2.2.1 Standardization of the translation of Chinese political discourse In the context of standardization of the translation of Chinese political discourse, the normative document "Basic Norms for the Translation of Chinese Political Discourse into English" (hereinafter referred to as the "Norms"), developed and published in 2023 by the Chinese Publishing House of Literature in Foreign Languages with the participation of a number of specialized institutions, is of particular importance. Despite the fact that this document is initially focused on English translation, its methodological value goes beyond a single language pair. The principles and approaches presented in the "Norms" are universal in nature and can be effectively applied when translating into Russian and other languages. The "Norms" are based on the fundamental principle that the translation of Chinese political discourse should provide "an optimal balance between the accuracy of the transmission of content and the naturalness of presentation for native speakers of the target language" [18, p. 4]. The implementation of this principle requires the translator not only a deep understanding of the political context and cultural characteristics of the source text, but also the ability to convey Chinese political concepts by linguistic means that correspond to the norms and traditions of the target language. The Norms are a comprehensive document covering the theoretical foundations and practical recommendations for the translation of political discourse. The document focuses on contextual adaptation and terminological accuracy, overcoming cross-cultural barriers and optimizing the workflow. These key aspects served as the theoretical foundation for the development of the MAGIC-PTF system, defining the main directions of its functioning and development. 2.2.2 Designing MAGIC-PTF based on "Norms" (1) Contextual adaptation and terminological accuracy The Stylist Agent plays a central role in ensuring contextual adaptation and terminological accuracy. Thanks to the use of fine-tuning technology [10], this component of the system effectively assimilates the stylistic features of Chinese political texts and their translations. Of particular importance is the integration into the system of a specially developed Chinese-Russian terminological corpus of political discourse, which ensures consistency and accuracy in the use of political terminology. (2) Overcoming cross-cultural barriers A key role in overcoming cross-cultural barriers is played by the Agent-Reader, who evaluates the text from the point of view of the Russian-speaking reader. He analyzes not only the accuracy of the transmission of culturally specific concepts, but also the conformity of the general style of presentation to the norms of the Russian language. Based on the comments received from the Reader Agent, the system consistently improves the translation, making it more understandable and natural for the Russian-speaking audience. (3) Workflow optimization The organization of the workflow in MAGIC-PTF fully complies with the recommendations of the "Norms", which provide for "primary translation, editing, initial verification, re-verification, unification of terminology and formatting, general viewing of the text, preparation of the final version" [18, p. 13]. In this process, each agent performs its own specialized function: the Stylist Agent creates the primary translation, the Translator Agent performs editing, the Editor Agent performs verification at various stages, the Reader Agent performs an overall assessment of the text, and the final version is formed by the Translator Agent. This interaction of agents, based on the principle of a professional translation team, ensures high quality of the final product. 3. Design and functional characteristics of the system agents 3.1 Stylist Agent The fundamental component of the MAGIC-PTF system is the Agent-Stylist, who, according to the "Norms", assumes the functions of the primary translator. Its main task is to create an initial version of the translation with a unified terminology and a consistent style. The work of this agent is based on the BYAM, which has undergone the procedure of retraining on a specially selected corpus of 500 parallel sentences extracted from the fourth volume of the book "On Public Administration" and its official translation into Russian. Retraining, being a key method of transfer learning, allows you to purposefully configure a pre-trained model using a specialized data set. This method is of particular importance in conditions of a limited amount of data, since it makes it possible to preserve the basic language competencies of the model while simultaneously acquiring specialized knowledge in a specific area. In the context of working with BYAM, the use of advanced training technology allowed MAGIC-PTF to effectively master both the stylistic features and the terminological apparatus of official translations of Chinese political discourse. This ensures the creation of a high-quality primary translation, which serves as the basis for further work within the framework of multi-agent interaction. 3.2 Agent-Translator The central place in the MAGIC-PTF system is occupied by a Translator Agent, who, in accordance with the "Norms", is responsible for editing the text and preparing its final version. Unlike the Stylist Agent, the functioning of the Translator Agent, as well as the Editor Agent and the Reader Agent, is based on industrial engineering technology, which ensures the most effective use of the BYAM capabilities. The main activity of the Translation Agent is to process the original Chinese text and the primary translation created by the Stylist Agent. At this stage, the agent applies his linguistic competence to improve the quality of translation, making it more natural and stylistically verified. In the future, based on the recommendations of the Agent-Editor and the Agent-Reader, a multi-stage revision of the text is carried out, which allows you to achieve the level of professional translation. 3.3 Agent Editor In the MAGIC-PTF structure, the Agent-Editor, according to the "Norms", performs primary and repeated verification of the translation, acting as a guarantor of its quality. His work covers a comprehensive analysis of the text from linguistic, semantic, cultural and political points of view. When receiving the original Chinese text and translation from the Translation Agent, the Agent-Editor primarily evaluates the accuracy and completeness of the information transfer. At the same time, special attention is paid not only to the content side, but also to the quality of linguistic expression. During the verification, a detailed analysis of grammatical structures, lexical composition and stylistic features of the text is carried out. Russian Russian translation Agent evaluates the compliance of the translation with the norms of the Russian language and usage, as well as the adequacy of the transfer of culturally specific concepts, taking into account their perception by native speakers of Russian culture. 3.4 Agent-Reader Within the framework of MAGIC-PTF, the Reader Agent implements the function of final reading of the text provided for by the "Norms". Its main task is to evaluate the translation from the point of view of a foreign—speaking reader, without referring to the original. The main attention of the Agent-Reader is paid to the expressiveness of the text and its perception by the target audience. At the same time, a number of key aspects are evaluated: how well the text meets the reader's expectations, whether it is able to interest the reader, elicit a response and stimulate further reflection. Special attention is paid to how adequately the cultural elements of the text are conveyed, taking into account the peculiarities of the Russian-speaking audience. An important component of the agent's work is the analysis of the potential perception of the text by the target audience. This reader's assessment provides important feedback for the final revision of the translation. 3.5 The process of interaction between intelligent agents The MAGIC-PTF is based on the interaction of four intelligent agents providing a comprehensive translation of Chinese political discourse into Russian. The translation process is implemented in stages: initially, the Stylist Agent forms a basic translation version, which is then pre-processed by the Translator Agent. At the next stage, the text undergoes a double check: The Agent-Editor carries out a professional examination of the translation, and the Agent-Reader evaluates it from the perspective of the target audience. Based on the recommendations received, the Translator Agent makes the necessary adjustments to the text. This iterative process continues until the required level of quality is achieved, after which the Agent-Translator forms the final version of the text and an analytical conclusion. Table 1 presents the scheme of MAGIC-PTF on the example of the translation of the following sentence from the eighth Chapter of the fourth volume: "必须看到, 全面建设社会主义现代化国家, 实现中华民族伟大复兴, 最艰巨最繁重的任务依然在农村, 最广泛最深厚的基础依然在农村." [19, c. 194]. Table 1. The scheme of interaction of intelligent agents
4. The results of the experiment 4.1 Analysis of transfers made by MAGIC-PTF As part of the analysis of the effectiveness of MAGIC-PTF, we will consider the results of the translation of two sentences from the fourth volume "On Public Administration" (see Table 2). When analyzing the first sentence, it should be noted the high level of bookishness of the MAGIC-PTF translation, which fully corresponds to the stylistic requirements of the political text. Special attention is given the precision of gear specific political terms: "全面建设社会主义现代化国家" (comprehensive construction of the modernized socialist state) and "实现中华民族伟大复兴" (the realization of the great rejuvenation of the Chinese nation) are fully consistent with the official translation. The achievement of the system was innovative solutions in the field of syntax and vocabulary. Splitting the original sentence into two components has significantly improved the readability of the text. The choice of the construction "Must be understood" instead of "Must be seen" more accurately conveys the communicative intention of the original. The use of the phrase "That is where our broadest and most solid foundation is located" demonstrates the effectiveness of the Reader Agent, providing natural coherence of the text and reducing the cognitive load on the reader. In translating the second sentence, the system also demonstrated a high level of competence. The book style of presentation has been preserved, which is achieved by using participial phrases and prepositive nominal constructions. The key political terms "五位一一" (five-fold) and "四个全全" (four comprehensive aspects) are accurately conveyed. It is noteworthy to use the construction "One should adhere to a systematic approach" for the translation of "要坚持系统观念", which, unlike the adverbial turnover in the official translation, retains the status of the main information of the sentence. Dividing the original sentence into two components also helps to improve the perception of the text. Table 2. Examples of transfers made by the MAGIC-PTF system
The comparative analysis of translations shows the significant advantages of the MAGIC-PTF system in the field of translation of political texts. During the research, it was revealed that this system demonstrates a high level of accuracy both in terms of stylistic design and in terms of terminological compliance with official documents. The system does not just copy the syntactic constructions of the original, but transforms them taking into account the norms of the Russian language, which makes the text more accessible to the Russian-speaking audience. It is important to note that such transformations do not affect the semantic accuracy of the translation. In the field of lexical design, the system demonstrates a deep understanding of the specifics of political discourse. This is manifested in the correct use of established political terms and the choice of language tools characteristic of the official business style of the Russian language. Moreover, MAGIC-PTF successfully copes with the transmission of subtle semantic nuances, which is especially important when translating political texts. The results obtained convincingly prove the effectiveness of the multi-agent approach in specialized translation and open up new prospects for the further development of automated translation systems for political texts. 4.2 Comparative analysis of the effectiveness of the MAGIC-PTF system with existing machine translation platforms To evaluate the effectiveness of the MAGIC-PTF system in the field of translating Chinese political discourse into Russian, a comprehensive experimental methodology was developed. The study included not only comparing MAGIC-PTF with existing machine translation systems, but also analyzing the effectiveness of its internal structure, as well as comparing it with other translation methods based on artificial intelligence. The experimental study was conducted in several directions. First of all, the MAGIC-PTF system was compared with the widely used DEEPL and ChatGPT-4o platforms, representing, respectively, traditional neural network machine translation and modern BYAM achievements in the field of translation. Further, to confirm the effectiveness of the multi-agent structure, the final MAGIC-PTF translation was compared with the original version generated by the Stylist Agent. Of particular interest was the study of the role of a specialized Stylist Agent: in a separate experiment, its functions were transferred to the ChatGPT-4o dialog mode while maintaining interaction with other agents of the system. Such a multidimensional approach to experimental research made it possible to obtain an objective assessment of both the MAGIC-PTF system as a whole and the effectiveness of its individual components in the context of translating political discourse. Table 3. Comparative analysis of various machine translation systems
The analysis of the translation results presented in Table 3 allows us to conclude about the advantages of the MAGIC-PTF system in the field of translating Chinese political discourse into Russian. At the same time, traditional machine translation systems have demonstrated a number of disadvantages. Thus, DEEPL translations reveal stylistic deviations and inaccuracies in the use of political terminology, as well as redundancy of lexical constructions. The ChatGPT-4o system, in turn, tends to unduly simplify the syntactic structure of the source text, which leads to distortion of the semantic shades of the original and a decrease in the overall quality of the translation. The highest results were achieved by the primary translation of the Stylist Agent, which demonstrated precise adherence to the stylistic norms of political discourse and the correct use of special terminology. It should be noted that the attempt to replace the Stylist Agent with ChatGPT-4o as the primary translation generator turned out to be ineffective: even while maintaining subsequent interaction with other agents of the system, it was not possible to achieve the desired level of stylistic compliance with official translations. This fact highlights the key role of a specially tuned Stylist Agent in ensuring high-quality translation of political texts. 4.3 Measuring the effectiveness of translation using the COMET metric To measure the effectiveness of MAGIC-PTF on an expanded corpus of texts, an experiment was conducted using the COMET metric (Crosslingual Optimized Metric for Evaluation of Translation) [22]. As a test material, 100 proposals were selected from two sources: the third volume of the book "On Public Administration" and the "Report on the Work of the Government of the People's Republic of China" for 2024, as well as their official translations into Russian. The choice of these texts is due to the need to test both the adaptive and generalization abilities of the model. The third volume "On Public Administration", belonging to the same series as the fourth volume used in MAGIC-PTF training, allows you to evaluate the effectiveness of the model on familiar material. In turn, the "Report on the work of the Government", covering a wide range of issues from the economy to the social sphere, differs significantly from the series "On Public Administration" both in content and style, which makes it possible to test the ability of the model to process heterogeneous material. As part of the experiment, four translation systems were comparatively tested: MAGIC-PTF, DEEPL, ChatGPT-4o and a variant using exclusively a Stylist Agent. The COMET metric was chosen as an evaluation tool due to a number of significant advantages over traditional metrics based on n-gram matching. Unlike metrics such as BLEU, COMET implements a deeper semantic analysis that allows you to evaluate not only lexical correspondences, but also the semantic equivalence of different ways of expressing the same thought. The system takes into account contextual information and demonstrates stable results when working with different language pairs, evaluating the quality of translation on a scale from 0 to 1. Of particular importance is the use of COMET in the analysis of the translation of political discourse, characterized by complex semantic structures and specific terminology. The ability of the metric to analyze deep semantic structures taking into account the context and effectively identify synonymous expressions allows for a comprehensive assessment of the quality of translation both from the point of view of preserving the original meaning and from the point of view of the adequacy of the chosen means of expression in the translation language. Table 4. COMET evaluation results for MAGIC-PTF and other machine translation systems
The analysis of the experimental results presented in Table 4 demonstrates the significant superiority of the MAGIC-PTF system over other translation systems under study. The model achieved the highest indicators according to the COMET metric both when working with the "Report on the Work of the Government" (0.8363) and with the third volume "On Public Administration" (0.8474), which indicates its high potential for generalization. It is significant that the system has not only successfully adapted to the specifics of specific texts, but has also demonstrated the ability to effectively identify and apply universal patterns of translation of Chinese political discourse into Russian. An interesting aspect of the study is the relatively low COMET score of the Stylist Agent, which is explained by the specifics of the metric itself, focused mainly on assessing semantic compliance without taking into account the stylistic parameters of the text. However, it is the Stylist Agent that forms the fundamental stylistic basis for the entire MAGIC-PTF system. A significant improvement in COMET indicators (by more than 6%) during the transition from the initial translation of the Stylist Agent to the final result of MAGIC-PTF indicates that the multi-agent interaction architecture effectively increases semantic accuracy and smoothness of presentation while maintaining the stylistic adequacy of the text. The superiority of MAGIC-PTF over DEEPL and ChatGPT-4o systems convincingly proves the effectiveness of integrating a specialized BAM retraining procedure with multi-agent interaction in the field of specialized translation. This approach provides higher translation quality compared to both universal BMS and traditional neural machine translation systems. It is noteworthy that MAGIC-PTF has achieved such results based on learning from a single text, which opens up prospects for further improving the effectiveness of the system by expanding the training corpus to include more diverse political texts. Conclusion In conclusion, it should be noted that the MAGIC-PTF system developed within the framework of this study is an innovative solution in the field of automated translation of Chinese political discourse into Russian. The architecture of the system, based on the interaction of four specialized agents, not only ensures full compliance with the requirements of the "Basic Norms of Translation of Chinese Political Discourse", but also significantly increases the adaptability and professional competence of BYAM in the field of translation of political texts. The effectiveness of MAGIC-PTF is based on three key components: specialized training of a Stylist Agent, which allows you to accurately reproduce the stylistic features and terminology of official translations; imitation of the process of teamwork of professional translators through the interaction of agents; and a multi-stage iteration mechanism that ensures continuous improvement in the quality of translation. Compared to traditional human translation, the system demonstrates a significant advantage in text processing speed and optimization capabilities. Thanks to multiple verification and adjustments within the framework of multi-agent interaction, the quality of MAGIC-PTF translation consistently approaches the level of professional translation, which is confirmed by high indicators of terminological accuracy, stylistic consistency and overall translation quality. The results of testing using COMET metrics convincingly demonstrate the superiority of MAGIC-PTF when working with various types of political texts, which indicates not only the successful development of the system of the specifics of translating specific texts, but also its ability to effectively identify and apply universal principles of translating Chinese political discourse into Russian. This achievement opens up new prospects for the use of BYAM in the field of professional translation and suggests that with the further development of technology and the accumulation of practical experience, such multi-agent translation systems will play an increasingly important role in the international broadcast of Chinese political discourse. References
1. Brown, T., Mann, B., Ryder, N., et al. (2020). Language models are few-shot learners. Advances in Neural Information Processing Systems, 33, 1877-1901.
2. OpenAI. (2023). GPT-4 technical report. arXiv preprint arXiv:2303.08774. 3. Zhou, J., Ke, P., Qiu, X., Huang, M., & Zhang, J. (2023). ChatGPT: potential, prospects, and limitations. Frontiers of Information Technology & Electronic Engineering, 1-6. 4. Hu, K., & Gao, L. (2024). Foreign Language Discipline Development in the Context of Large Language Models: Problems and Prospects [Развитие иностранных языков в контексте больших языковых моделей: проблемы и перспективы]. Foreign Language World, 2, 7-12. 5. Qin, H., & Zhou, X. (2024). Large Language Models and Comparative Language Studies [Большие языковые модели и сопоставительное изучение языков]. Foreign Language Teaching and Research, 2, 163-176. 6. Hu, K., & Li, X. (2023). Development of Translation Studies in the Context of Large Language Models: Problems and Prospects [Развитие переводческих исследований в контексте больших языковых моделей: проблемы и перспективы]. Chinese Translators Journal, 6, 64-73. 7. Feng, Q., & Zhang, K. (2024). Analysis of AI-assisted Foreign Language Teaching and Research Capabilities: Taking ChatGPT-4 and Wenxin Large Model 4.0 as Examples [Анализ возможностей искусственного интеллекта в преподавании иностранных языков и исследованиях]. Computer-Assisted Foreign Language Education, 3, 3-12. 8. Wang, H., & Xie, F. (2024). Research on Innovation of Translation Education Practice Model Driven by Large Language Model Technology [Инновационное исследование практической модели переводческого образования]. Chinese Translators Journal, 2, 70-78. 9. Su, Q. (2024). Analysis of the Application Effectiveness of Large Language Models in Second Language Teaching [Анализ эффективности применения больших языковых моделей в преподавании второго языка]. Foreign Language World, 3, 35-42. 10. Zhang, B., Haddow, B., & Birch, A. (2023). Prompting large language model for machine translation: A case study. In International Conference on Machine Learning, PMLR (pp. 41092-41110). 11. Zhang, X., Rajabi, N., Duh, K., & Koehn, P. (2023). Machine translation with large language models: Prompting, few-shot learning, and fine-tuning with QLoRA. Proceedings of the Eighth Conference on Machine Translation, 468-481. 12. Moslem, Y., Haque, R., Kelleher, J. D., & Way, A. (2023). Adaptive machine translation with large language models. arXiv preprint arXiv:2301.13294. 13. Li, J., Zhou, H., Huang, S., Cheng, S., & Chen, J. (2024). Eliciting the translation ability of large language models via multilingual finetuning with translation instructions. Transactions of the Association for Computational Linguistics, 12, 576-592. 14. Du, Y., Li, S., Torralba, A., Tenenbaum, J. B., & Mordatch, I. (2023). Improving Factuality and Reasoning in Language Models through Multiagent Debate. arXiv preprint arXiv:2305.14325. 15. Wu, Q., Bansal, G., Zhang, J., et al. (2023). Enabling Next-Gen LLM Applications via Multi-Agent Conversation Framework. arXiv preprint arXiv:2308.08155. 16. OpenAI. (2024). ChatGPT (version June 20) [Large language model]. Retrieved from https://chat.openai.com/chat 17. Andreas, J. (2022). Language Models as Agent Models. Findings of the Association for Computational Linguistics: EMNLP, 5769-5779. 18. Editorial Group (Ed.). (2023). Basic Standards for Translating Chinese Political Discourse into English. Beijing: Foreign Languages Press. 19. Xi, J. (2022). O gosudarstvennom upravlenii (Vol. 4) [In Chinese]. Beijing: Foreign Languages Press. 20. Xi, J. (2023). O gosudarstvennom upravlenii IV [Translated from Chinese]. Beijing: Foreign Languages Press. 21. DeepL. (2024). DeepL Translator [Online translation tool]. Retrieved from https://www.deepl.com/zh/translator#zh/ru/ 22. Guerreiro, N. M., Rei, R., Stigt, D., et al. (2024). xComet: Transparent Machine Translation Evaluation through Fine-grained Error Detection. Transactions of the Association for Computational Linguistics, 12, 979-995.
Peer Review
Peer reviewers' evaluations remain confidential and are not disclosed to the public. Only external reviews, authorized for publication by the article's author(s), are made public. Typically, these final reviews are conducted after the manuscript's revision. Adhering to our double-blind review policy, the reviewer's identity is kept confidential.
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
© 1998 – 2025 Nota Bene. Publishing Technologies. NB-Media Ltd.
