Earticle

This study is a comparative analysis of translated-text error-rates in song lyrics by K-Pop group New Jeans. Seven non-official machine translations (MTs) of ten songs were analyzed against official human-translated lyrics. The ten songs were consisted of a total 235 segments and the seven MTs were categorized under neural-network types (DeepL, Papago, Google Translate) and generative-AI types (ChatGPT, Bard, ClovaX, MS Bing Translate). Analysis discovered three salient points. First, neural-network types presented significantly higher error rates than generative-AI types. DeepL (66%), Papago (64%), Google Translate (59%). The most common errors were semantic and grammatical. A common feature of the errors was the poor contextual understanding and consistency between consecutive segments. This suggests that neural network MTs may have limited application for translating K-pop lyrics, which are expressive text. Second, neural network MTs were twice as erroneous as generative AI translations, with the official human translation as the baseline. T his suggests that AI translation may be more useful for translating K -pop lyrics than neural network MT in terms of semantic accuracy an d structural form. Third, generative AI translation quality improved in general when additional parameters and descriptions were provided via the services’ chat functions.

In this study, we carried out a series of experiments to explore how ChatGPT (version 4o) evaluated Korean-English translations. Using two datasets of human translations (n=57) and two datasets of post-edited translations (n=56), all drawn from Lee and Lee (2021), we adopted two evaluation approaches with strict prompt control. In Experiment A, ChatGPT rated the four datasets freely on a five-point scale without specific criteria. In Experiment B, which was conducted concurrently with Experiment A, ChatGPT rated the same datasets using a prescribed, criterion-referenced five-point scale. To assess intra-rater reliability, we repeated both experiments one month later. This study yielded both quantitative and qualitative findings, including the following: (1) ChatGPT’s average scores differed significantly from those of human raters; (2) correlations between human and ChatGPT scores ranged from ‘moderate’ to ‘strong’; (3) the use of the prescribed rating scale improved ChatGPT’s reliability as a rater; (4) ChatGPT exhibited very low intra-rater reliability; and (5) ChatGPT’s self-justifications for its ratings varied in quality, often failing to identify obvious errors.

This study presents the intermediate results of an ongoing action research project on a business interpreting class model using ChatGPT. The focus is on providing structured guidance for the acquisition of domain-specific knowledge and interpreting skills in the domain of business interpreting. To address the issue of ‘hallucination’ in ChatGPT, the model used an external domain-specific textbook as course material and input data for the generative AI prompts. The resulting AI-generated role-play scripts were used in weekly lectures at a graduate-school in Seoul. Thirty-six students in four classes took turns interpreting in the new class design. The following research questions were addressed through literature review, output generated using prompts, anonymous post-training surveys, and instructor observation notes: (1) How can a class model for business interpreting practice be designed and implemented using ChatGPT? (2) What are the students’ responses? (3) How can the class model be improved further? Various class data and anonymous post-training survey data were analyzed to gain insight into the learners’ experience with the new class model and how it may be enhanced for future iterations.

This study compares how Papago, Google Translate, and ChatGPT perform when translating police dialogues, using BLEU scores, manual assessment, and error analysis. The analysis results are as follows. All three tools showed low BLEU scores compared to reference translations. In manual evaluation, ChatGPT significantly outperformed others, scoring 8.6 out of 10, compared to Google Translate's 6.5 and Papago's 5.6. Error analysis confirmed ChatGPT's superiority, with only 41 errors, while Google Translate and Papago produced 123 and 152 errors respectively. Across all tools, substitution was the most common accuracy error, followed by omission and addition. These results suggest that ChatGPT is the most reliable tool for police communication with foreign nationals.

This study compares the English and Japanese translations of The Vegetarian (Part I) by Han Kang, focusing on how translation strategies differ according to high-context and low-context cultural frameworks. The analysis centers on three scenes—Yeong-hye’s physical description, her husband’s self-characterization, and the “Dreams of murder” passage—to identify how each translation negotiates cultural and interpretive shifts. In the English version, the translator employs a range of explicitation strategies to accommodate low-context readers, manifested in four main types: explicitation with deletion, judgmental insertion, mistranslation, and addition. These strategies often restructure the original’s implicit tone, emotional ambiguity, and narrative restraint into clarified and interpretively guided expressions, leading to semantic and stylistic shifts. In contrast, the Japanese translation demonstrates high fidelity, preserving the original’s indirectness, lexical nuance, and affective texture in line with high-context communication norms. This comparative analysis shows that literary translation is not merely linguistic transfer but a culturally embedded act of interpretive reconstruction. The study concludes by emphasizing the pedagogical implications of training translators to recognize contextual asymmetries, navigate the ethical boundaries of interpretation, and maintain the tonal integrity of high-context narratives.

This study examines how collaborative translation has been approached in Korean scholarship and explores its conceptual framework and potential for development. Unlike in international research, where the term is more clearly defined, collaborative translation remains underused and insufficiently theorized in Korea. To address this gap, the study analyzes 254 academic papers to identify prevailing trends and limitations. Findings show that machine-centered approaches dominate (61.4%), focusing mainly on MT evaluation and translation education, while studies on professional collaboration between human translators and MT systems are relatively scarce. Human-centered studies (38%) encompass both multi-translator models—such as online collaboration, relay translation, and co-translation—and single-translator models involving collaboration with editors, proofreaders, clients, and others. Despite the widespread practice of co-translation, it remains underexplored in academic literature, exposing a gap between industry and academia. Existing classification models also fail to account for cases where a single translator assumes multiple roles, underscoring the need for a more nuanced framework that considers the intensity, stages, and dynamics of collaboration. This study highlights the need for a clearer and more comprehensive conceptualization of collaborative translation, and calls for future research that bridges theoretical insights with evolving technological and professional practices.

This study examines the inherent limitations of Artificial Intelligence Translation (AIT) through the lens of universal grammar and context-oriented translation theories, with particular reference to Wittgensteinian philosophy of language. Through analysis of translation errors in Russian-Korean AIT outputs, this study argues that certain limitations stem from fundamental algorithmic constraints rather than merely technical implementation challenges. While AIT systems effectively leverage vector databases to store, manage, and optimize word relationships in context, allowing for increasingly fluid translations, their operation remains fundamentally aligned with Wittgenstein's concept of language games. However, significant translation errors persist in pragmatic dimensions, including narrator intention, idiomatic expressions, nuance, and rhetorical connotations. In essence, AIT systems struggle to capture what Benveniste terms "language outside the text" - the domain of enunciation. Despite their considerable capabilities, AIT systems remain confined to text-to-text operations, unable to fully engage with the broader space of communication between narrators and auditors where idioms and significations are dynamically generated. While AIT can translate the narrative, it seems unable to grasp narration itself; it can translate utterances but not the act of enunciation - a capability that remains uniquely human.

With the popular uptake of ChatGPT, the interpreter community is set to undergo significant changes in its discussion and directions on topics related to machine translation. This study investigates the features and differences of ChatGPT, a natural language processing (NLP) tool, in comparison to Google Translate’s translation engine, specifically between the Korean and Japanese language pair, with the purpose of exploring means to leverage ChatGPT’s capabilities in the context of interpreter training in said pair. For this purpose, we will first analyze ChatGPT’s translation of idioms and proverbs to identify whether it is capable of paraphrasing polysemous connotations and implications. We will next analyze ChatGPT’s ability to translate unstructured colloquial utterances. Finally, we will analyze whether ChatGPT’s translation output displays translation universals such as simplification and clarification. Based on these analyses, this study will discuss how ChatGPT may be used toward interpreter training.

This study aims to evaluate the quality of AI translation in comparison to human translation, identify challenges in applying metaphor translation methods to AI-generated texts, and examine the relationship between translation quality and the strategies employed. The analysis delves into the Korean translation of metaphorical expressions from the Chinese literary work Dawn Blossoms Plucked at Dusk, utilizing outputs from Google Translate and ChatGPT. The initial step of this study involved the categorization and statistical analysis of translation errors generated by AI. The findings show that literal translations and omissions were relatively infrequent, whereas distortions of meaning and stylistic issues were more prevalent. Furthermore, this study revealed that the types and frequency of errors were reduced when AI translation employed the same methods as human translation. It is anticipated that the findings of this study will serve as foundational data for future research. However, the study is constrained by limited data and a lack of diversity, which calls for further investigation to overcome these limitations.

As K-Webtoons gain global popularity, global fans captivated by K-Webtoons ask for more content to be translated faster than before. In webtoons, onomatopoeia and mimetic words play a crucial role in creating vivid imagery and strengthening sensory effect, so studies on how to translate onomatopoeia/mimetic words in webtoons would contribute to the globalization of the K-Webtoons. Previous research on onomatopoeia translation has mostly focused on onomatopoeia/mimetic words in literary texts such as novels or poetry. Onomatopoeia/mimetic word translation in webtoon is quite different from that in literary texts; therefore, this study first investigates how webtoon onomatopoeia/mimetic words are translated into English. Second, as the demand for webtoon translation rises, faster translation is also getting important. To satisfy this rising demand, it is explored to utilize machine translation in webtoon translation. Webtoon machine translation outcome reveals that it has limitations in translating polysemous onomatopoeia/mimetic words. Hence, in the final chapter, the concept and development of MMT is introduced as a way of complementing the limitation of existing machine translation and specific ways to enhance the performance of MMT are suggested.