Machine translation in language education: Perspectives from advanced language learners

The recent advent of more powerful machine translation (MT) tools has significant implications for foreign language teaching and learning. While current research and pedagogical guidelines on MT in the foreign and second language classroom have focused primarily on survey data, less is known about how students actually use MT tools. The current article chronicles a computer tracking study of novice learners of French as a Foreign Language (n=26) as a way to inform this field of research. Drawing on diverse data sources (screen-recorded observations, retrospective recalls, post-interviews) and a Critical Incident Technique analytical approach, the article showcases the actions and cognitive processes related to MT that both supported and hindered student participants in their foreign language writing. Supportive behaviors and mindsets centered on a specific awareness of MT tool limitations paired with appropriate action while detrimental behaviors and mindsets centered on inappropriate input, lack of output analysis, and time. Suggestions for how these use-based findings might inform how MT is discussed in the language classroom conclude the article.

The present study seeks to address the emotional lexical units in the context of Turkish synonyms that fall under the categories of terms for pain, joy, and excitement, in addition to examining potential English equivalents based on two machine translation tools: Google Translate and DeepL. Additionally, the researchers sought to identify whether two distinct machine translation systems presented near-synonyms in the same way, whether machine translation algorithms considered context when translating emotive lexical units, and how DeepL and Google Translate translated emotive lexical units differently. In this case, Google Translate and DeepL were used to assess sentences from the TDK online website that covered each category and had a total word count of 25. The findings have demonstrated that the machine translation systems DeepL and Google Translate typically produce translations that are comparable, that Google Translate platform considers context more when translating synonyms, and that Google Translate also considers words that are specific to Turkish or words borrowed from Arabic, Persian, and Ottoman Turkish, whereas DeepL occasionally failed in this area. Another finding is that machine translation tools have advanced to a great level, considering near-synonym translation. It is recommended that scholars benefit from the near-synonym corpora gathered in other subjects and their detections in other language pairs and that future research assess synonyms in other fields using machine translation or in various settings.

This study investigated the impact of an intervention utilizing ChatGPT in literary translation on EFL student translators’ comfort level, proficiency in machine translation (MT) tools, openness to embrace AI technology, and self-efficacy in literary translation. Additionally, the study aimed to identify the benefits and limitations reported by the participants in using ChatGPT for literary translation. Using an experimental pre-post design and purposive non-probability sampling, 50 student translators participated in the study. Data collection instruments included a pre-post perceived self-efficacy questionnaire and follow-up semi-structured interviews. Quantitative data analysis indicated the effectiveness of the study intervention in enhancing student translators’ proficiency in using MT tools and their comfort level with them. Significant differences were observed in student translators’ self-efficacy beliefs before and after the intervention. Additional analysis demonstrated a significant difference in student translators’ confidence before and after the intervention, accompanied by a medium effect size. The analysis of participants’ responses to the questionnaire open questions revealed notable benefits and limitations associated with using ChatGPT in literary translation. Furthermore, the qualitative analysis of the interview data yielded valuable insights into the transformative effect of the ChatGPT intervention on students’ openness, comfort and confidence in using AI tools in literary translation.

The objective of the research is to review the ability of online machine translator tools includes Google Translate (GT), Collin Translator (CT), Bing Translator (BT), Yandex Translator (YT), Systran Translate (ST), and IBM Translator (IT). This research applies descriptive qualitative. The documentation was used in this study. The result of the analysis shows that the translation results are different, both from the style of language and the choice of words used by each machine translation tool. Thus, directly or indirectly, whether consciously or not, each translation machine carries its characteristics. Machine translation technology cannot be separated from the active role of humans. In other words, it will always be the best choice for users to rely on expert translation rather than machine translation. But no machine translator can be as accurate as human skills in producing translation products. In particular, the field of translation is also concerned with machine translation to support the performance of translators in analyzing the diction used as an element of language. In this regard, it needs to be underlined that the existence of machine translation is an additional facility in the world of translation, not as the main means of translation because the sophistication of the machine will not be able to match the flexibility of the human brain's cognitive abilities in adjusting the translation results according to the existing context. Accurate translation is sometimes subjective, relatively often temporal. Therefore, it is permissible for translating by more than one machine translator

Machine Translation (MT) tools have advanced to a level of reliability such that it is now opportune to consider their place in language teaching and learning. Given their potential, the current study sought to engage EFL university sophomores in recursive editing afforded by Google Translate (GT) for one semester, and investigated (1) whether the students were able to correct errors assisted by GT, (2) whether GT facilitated better writing, (3) which aspects of writing, fluency, complexity or accuracy GT better assisted the learners with, and (4) the student’s attitudes toward GT. A quasi-experimental approach was adopted where the experimental group, comprising 34 students, received training in recursive editing while the control group, comprising 33 students, did not. Both groups completed an error-correction test and an essay pre- and post-intervention. The experimental group completed an additional Affective Filter and Task Difficulty questionnaire post-intervention. The results showed that the experimental group significantly outperformed the control group on the error correction test; however, no significant difference in the two groups’ writing scores was found. Further, the evaluation of writing fluency, accuracy, and complexity yielded a mixed picture. The control group demonstrated significantly better fluency as shown in sentence, clause, and T-unit counts, but not in total word count. The experimental group, conversely, showed significantly higher syntactic complexity in mean length of clause, and significantly better accuracy in writing post-intervention. Finally, the students were generally positive about using GT to learn in terms of anxiety, motivational belief, and task complexity.

Artificial Intelligence or AI as it is popularly known and its corollary, Machine Translation (MT) have long engaged scientists, thinkers and linguists alike in the twenty first century. However, the wider question that lies in the relation between technology and translation is, What does technology do to language? This is an important question in the current paradigm because new translation technologies, such as, translation memories, data-based machine translation, and collaborative translation, far from being just additional tools, are changing the very nature of the translators’ cognitive activity, social relations, and professional standing. In fact, in some translation situations such as when translating technical materials or subject matter that are not a specialization with human translators, one potentially needs technology. The purview of this paper, however, is limited to the role of MT in day to day situations where the generic MT tools like Google Translate or Bing Translator are encouraged. Further, it endeavours to weigh and empirically demonstrate the pros and cons of MT with a view to recommending measures for better communication training in the EFL set up of Saudi Arabia. Keywords: AI, MT, translation, technology, necessity, communication

Focusing on the topic of applying machine translation (MT) to Chinese–English subtitling, the subject of the present study, this article first analyses the current literature and introduces the general constraints in Chinese–English subtitling from three perspectives: technical, cultural, and textual. Technical constraints are imposed by the limited time and space available for each subtitle; cultural constraints relate to the disparities in the beliefs, values, customs, behaviours, and artefacts of different cultural groups; and textual constraints predominantly manifest in the differences between the source and target languages and the segmented nature of subtitles. In order to respond to these constraints and to achieve conciseness, comprehensibility, and coherence in the translated subtitles, this study highlighted condensation, context, and coordination as the key strategies to adopt. However, these strategies pose considerable challenges for MT in Chinese–English subtitling. First, the common practice of full translation by most MT tools tends to work against making subtitles concise. Second, the lack of relevant contextual knowledge limits the ability of MT to generate appropriate translations. Third, the segmented display of subtitles makes it challenging for MT to capture and reflect the logic in the source text, on the one hand, and to produce coherent output in self-contained segments, on the other. To illustrate these constraints and challenges, this article provides examples of bilingual subtitles from American Factory (Bognar & Reichert, 2019), an Oscar Winner for Best Documentary Feature, and compares the official subtitles with those generated by three popular MT tools in Chinese–English translation. The study investigated the efficacy of MT subtitling and its potential to produce quality subtitles; and this article proposes possible solutions and suggestions for improving the quality of MT in subtitling.

ABSTRACT: This study aims to analyze how machine translation tools, such as Google Translate and DeepL, translate religious terms on the Indonesia Travel website, a tourism platform rich in cultural and religious elements. This research employed a qualitative method, examining 10 Indonesian English articles and identifying 15 representative religious terms based on AlGhamdi's (2016) classification. The term is translated using both Google Translate and DeepL, and the results are analyzed using Newmark's (1988) translation procedure framework. The findings showed that both machines apply six procedures in handling religious terms, with transference and recognized translation being the most frequently used. However, there are differences in how Google Translate and DeepL strike a balance between accuracy and readability. Google Translate simplifies translations, improving accessibility but risking the loss of cultural nuance. In contrast, Google Translate preserves more of the religious and cultural depth, although this can reduce clarity for general readers. In conclusion, while both machines are effective in handling recognized religious terms, Google Translate is considered more effective for the Indonesia Travel website, as it delivers more transparent and more accessible translations for global audiences. However, post-editing is still recommended to ensure that translations are both accurate and readable.

We present an approach called MTP (multiple translation paths) aiming at assisting human translation in SNOMED CT localisation projects based on free, web-based machine translation tools. For a chosen target language, MTP generates a scored output of translation candidates (TCs) for each input concept. This paper describes the basic idea of MTP, the distribution of its output TCs and discusses typical examples with German as target language. The MTP approach capitalises on combinatorial growth by the combination of input languages, support languages, and translation engines. We applied MTP on the SNOMED CT Starter Set, using Google Translator, DeepL and Systran, together with the four source languages English, Spanish, Swedish and French, and Danish, Dutch, Norwegian, Italian, Portuguese, Polish and Russian as support languages. The descriptive assessment of TC variety, together with an analysis of typical results is the focus of this paper. MTP defines, for each input concept, TPs by the combination of input languages, support languages and translation engines, resulting in 91 translation results with various degrees of co-incidence (cardinality). The most configurations produce an average number of TCs indicating that the same TC is often derived via different translation paths. Combinations of translation engines result in distributions with a higher number of distinct TCs per concept. We present work in progress on using machine translation (MT) for terminology translation, by leveraging several free MT tools fed by different languages and language combinations. A first qualitative analysis was promising and supports our hypothesis that a majority voting applied to many translation candidates yields higher quality results than from one single engine and input language.

This paper presents a comparative analysis between human translation and automated outputs within the field of Turkish-Italian lexicography. Using a corpus of example sentences extracted from Turklex—a developing online multilingual dictionary—translations were carried out manually by a professional translator and subsequently generated by four different machine translation tools: Google Translate, DeepL, Gemini, and ChatGPT-4. Particular attention is paid to the rendering of headwords, a critical element in lexicographic accuracy. By examining the divergences and overlaps across these translations, the study highlights the strengths and limitations of current NMTs and AI-driven chatbots when applied to a less commonly examined language pair. To the best of our knowledge, this is the first focused investigation into Turkish-Italian translation within a lexicographical framework. The findings aim to inform both the evaluation of machine translation tools and the broader discussion on their integration into language resource development. The research found that while ChatGPT slightly outperformed the other tools, AI chatbots and NMTs still fall short of human translators in delivering translations that are both contextually nuanced and semantically accurate.

As artificial intelligence (AI) continues to redefine the boundaries of linguistic research, this study examines the extent to which machine translation (MT) and AI tools can go beyond literal meaning, push beyond surface-level syntax and semantics to process context-sensitive issues in literary translation. While traditional MT systems such as Google Translate and Microsoft Translator are optimized for direct source-to-target mapping, AI language models like ChatGPT represent a broader category of tools designed for general-purpose language generation, including but not limited to translation. Using a 14-line Arabic poem, translations were generated by three MT systems, one AI model (ChatGPT), and two certified human translators. These outputs were evaluated against ten linguistic and stylistic dimensions: punctuation, layout, rhyme, mood, theme, logico-semantics, transitivity, field, tenor, and mode. The six translation versions were compared using a framework grounded in systemic functional linguistics (SFL). The analysis also considers how humans process cognitive-linguistic features when rendering poetic language. Results indicate that ChatGPT outperformed both MT systems and human translators in structural and semantic coherence, as well as in preserving poetic features such as rhyme and mood. However, all automated systems struggled with context-rich dimensions like tenor and mode, underscoring the enduring value of human interpretive depth. The findings highlight the potential of AI language models to complement, rather than replace, human expertise in literary translation and advocate for hybrid approaches that integrate computational efficiency with poetic language and cultural sensitivity. Acknowledgements We would like to express our sincere appreciation to the human translators who contributed to this study by translating the selected poem from Arabic into English. We equally thank the two translation experts who contributed to the analysis of translations. Their careful attention to linguistic nuance and cultural context played a significant role in comparing the machine-generated and human-produced translations. The authors are thankful to the Deanship of Graduate Studies and Scientific Research at the University of Bisha for supporting this work through the Fast-Track Research Support Program. Disclosure Statement The authors reported no potential conflict of interests.

Modeling workflow to design machine translation applications for public health practice

Machine Translation (MT) should be integrated into Master of Translation and Interpreting (MTI) translation teaching more intelligently. This study focuses on the practical application of MT in MTI translation teaching, aiming to overcome challenges such as the rigidity in traditional teaching models, the monolithic methods for translation quality assessment, and the shortage of qualified teachers. Through an analysis of the current status of MTI translation teaching, this paper proposes strategies like utilizing MT for multi-modal teaching, conducting post-translation quality assessment training with MT, and implementing teacher training with MT tools. The results indicate that these strategies not only significantly improve teaching quality and efficiency but also promote innovation in teaching models. This paper offers novel insights into the MTI translation teaching method and provides solid support for the deep integration of MT with educational practice, thereby ushering in an important pathway for cultivating outstanding translation talents in the new era.

The web as corpus is considered helpful in translation training, although the literature argues that Internet data might be too chaotic and badly organised.For this reason, web concordancers are often resorted to in order to carry out web searches as they are more linguistically oriented and user-friendly.At the same time, machine translation (MT) has made huge strides forward in the last decades, and scholars have increasingly focused on MT in technical translation training.However, so far little has been investigated on the quality of MT vis--vis translations via web concordancers.This paper wishes to bridge this gap and aims at exploring whether and to what extent the web-based translation of a medical article carried out by consulting the Leeds web-concordancer is qualitatively more satisfactory than a translation using MT, or vice-versa.To this aim, a classroom observation study is undertaken with 9 Italian final-year bachelor's students.The students' Leeds-driven translations are compared with the automatic translations performed by two MT tools (DeepL and MateCaT).The paper findings highlight the overall correctness of the machine-translated sentence structures as well as of the related grammar.However, MT presents lexical and collocational issues mostly due to infrequent word usages.By contrast, the web-driven students' translations present a wider variety of lexical and collocational choices, despite some grammatical shortcomings.The paper suggests further research to investigate whether more training in translations based on web concordancers can help students feel at ease with online searches and reduce inaccuracies.

This book explains the concept, framework, implementation, and evaluation of controlled document authoring in this age of translation technologies. Machine translation (MT) is routinely used in many situations, by companies, governments, and individuals. Despite recent advances, MT tools are still known to be imperfect, sometimes producing critical errors. To enhance the performance of MT, researchers and language practitioners have developed controlled languages that impose restrictions on the form or length of the source-language text. However, a fundamental, persisting problem is that both current MT systems and controlled languages deal only with the sentence as the unit of processing. To be effective, controlled languages must be contextualised at the document level, consequently enabling MT to generate outputs appropriate for their functional context within the target document. With a specific focus on Japanese municipal documents, this book establishes a framework for controlled document authoring by integrating various research strands including document formalisation, controlled language, and terminology management. It then presents the development and evaluation of an authoring support system, MuTUAL, that is designed to help non-professional writers create well-organised documents that are both readable and translatable. The book provides useful insights for researchers and practitioners interested in translation technology, technical writing, and natural language processing applications.