Translation Task Research Articles

Structure and Intensity Unbiased Translation for 2D Medical Image Segmentation.

Data distribution gaps often pose significant challenges to the use of deep segmentation models. However, retraining models for each distribution is expensive and time-consuming. In clinical contexts, device-embedded algorithms and networks, typically unretrainable and unaccessable post-manufacture, exacerbate this issue. Generative translation methods offer a solution to mitigate the gap by transferring data across domains. However, existing methods mainly focus on intensity distributions while ignoring the gaps due to structure disparities. In this paper, we formulate a new image-to-image translation task to reduce structural gaps. We propose a simple, yet powerful Structure-Unbiased Adversarial (SUA) network which accounts for both intensity and structural differences between the training and test sets for segmentation. It consists of a spatial transformation block followed by an intensity distribution rendering module. The spatial transformation block is proposed to reduce the structural gaps between the two images. The intensity distribution rendering module then renders the deformed structure to an image with the target intensity distribution. Experimental results show that the proposed SUA method has the capability to transfer both intensity distribution and structural content between multiple pairs of datasets and is superior to prior arts in closing the gaps for improving segmentation.

Correcting translation for non-autoregressive transformer

Non-Autoregressive Transformer has shown great success in recent years. It generally employs the encoder–decoder framework, where the encoder maps the sentence into hidden representation, and the decoder generates the target tokens simultaneously. Since the theory of non-autoregressive transformer is consistent with the architecture of the encoder, we suppose that it is somewhat wasteful for the encoder to only map input sentence into hidden representation. In this study, we proposed a novel non-autoregressive transformer to fully exploit the capabilities of the encoder. Specifically, in our approach, the encoder not only encodes the input sentence into hidden representation, but also generates the target tokens. Consequently, the decoder is relieved of its responsibility to generate the target tokens, instead of focusing on correcting the sentence produced by the encoder. We evaluate the performance of the proposed non-autoregressive transformer on three widely-used translation tasks. The experimental results illustrate the proposed method can significantly improve the performance of the non-autoregressive transformer , which achieved 27.94 BLEU on WMT14 EN → DE task, 33.96 BLEU on WMT16 EN → RO task, and 33.85 BLEU on IWSLT14 DE → EN.

Developing Semantic Textual Similarity for Guragigna Language Using Deep Learning Approach

Semantic Similarity is one of the highest levels of NLP. STS has significant advantages in NLP applications like information retrieval, information extraction, text summarization, data mining, machine translation, and other tasks. This research aims to present a deep learning approach for capturing semantic textual similarity (STS) in the Guragigna. The methodology involves collecting a Guragigna language corpus and preprocessing the text data and text representation is done using the Universal Sentence Encoder (USE), along with word embedding techniques including Word2Vec and GloVe and mean Square Error (MSE) is used to measure the performance. In the experimentation phase, models like LSTM, Bidirectional RNN, GRU, and Stacked RNN are trained and evaluated using different embedding techniques. The results demonstrate the efficacy of the developed models in capturing semantic textual similarity in the Guragigna language. Across different embedding techniques, including Word2Vec, GloVe, and USE, the Bidirectional RNN model with USE embedding achieves the lowest MSE of 0.0950 and the highest accuracy of 0.9244. GloVe and Word2Vec embedding also show competitive performance with slightly higher MSE and lower accuracy. The Universal Sentence Encoder consistently emerges as the top-performing embedding across all RNN architectures. The research results demonstrate the effectiveness of LSTM, GRU, Bi RNN, and Stacked RNN models in measuring semantic textual similarity in the Guragigna language.

MV2MV: Multi-View Image Translation via View-Consistent Diffusion Models

Image translation has various applications in computer graphics and computer vision, aiming to transfer images from one domain to another. Thanks to the excellent generation capability of diffusion models, recent single-view image translation methods achieve realistic results. However, directly applying diffusion models for multi-view image translation remains challenging for two major obstacles: the need for paired training data and the limited view consistency. To overcome the obstacles, we present a first unified multi-view image to multi-view image translation framework based on diffusion models, called MV2MV. Firstly, we propose a novel self-supervised training strategy that exploits the success of off-the-shelf single-view image translators and the 3D Gaussian Splatting (3DGS) technique to generate pseudo ground truths as supervisory signals, leading to enhanced consistency and fine details. Additionally, we propose a latent multi-view consistency block, which utilizes the latent-3DGS as the underlying 3D representation to facilitate information exchange across multi-view images and inject 3D prior into the diffusion model to enforce consistency. Finally, our approach simultaneously optimizes the diffusion model and 3DGS to achieve a better trade-off between consistency and realism. Extensive experiments across various translation tasks demonstrate that MV2MV outperforms task-specific specialists in both quantitative and qualitative.

Ancient Text Translation Model Optimized with GujiBERT and Entropy-SkipBERT

To cope with the challenges posed by the complex linguistic structure and lexical polysemy in ancient texts, this study proposes a two-stage translation model. First, we combine GujiBERT, GCN, and LSTM to categorize ancient texts into historical and non-historical categories. This categorization lays the foundation for the subsequent translation task. To improve the efficiency of word vector generation and reduce the limitations of the traditional Word2Vec model, we integrated the entropy weight method in the hopping lattice training process and spliced the word vectors with GujiBERT. This improved method improves the efficiency of word vector generation and enhances the model’s ability to accurately represent lexical polysemy and grammatical structure in ancient documents through dependency weighting. In training the translation model, we used a different dataset for each text category, significantly improving the translation accuracy. Experimental results show that our categorization model improves the accuracy by 5% compared to GujiBERT. In contrast, the Entropy-SkipBERT improves the BLEU scores by 0.7 and 0.4 on historical and non-historical datasets. Ultimately, the proposed two-stage model improves the BLEU scores by 2.7 over the baseline model.

Hokhmat ha-Nefesh : Soul and Dream, from Eleazar of Worms to Renaissance Italy

Abstract This article examines the traditions of copying and reading a medieval Ashkenazi treatise, Hokhmat ha-Nefesh (»Wisdom of the Soul«) by Eleazar of Worms (1176–1238) in Renaissance Rome. It delves into a manuscript copied in 1515 by Jewish scholar Elias Levita (1469–1549), commissioned by Levita’s patron, Neoplatonist and Hermetic scholar Egidio da Viterbo (1469–1532). In addition to copying the text, Egidio had Jewish interpreters translate Hokhmat ha-Nefesh into vernacular Italian. The treatise discusses the heavenly nature of the soul, its connections with the angelic world, its prophetic abilities, and its fate and retribution in the afterlife. What role did this Ashkenazi text play in the early modern tradition? How can this text be seen as a testament to kabbalistic perspectives in Renaissance Italy? How did Egidio’s collaborators handle their scribal and translation tasks, and the relationship between the Hebrew original and the vernacular translation of Hokhmat ha-Nefesh? This essay addresses these questions.

Los fraseologismos en el aula de prácticas de traducción: estudio de caso sobre el impacto de los traductores automáticos

Abstract The use of machine translation (MT) systems by undergraduate and postgraduate students has been analysed in different studies (Sycz-Opoń and Gałuskina 2017; Jia et al. 2019; Loock and Léchauguette 2021, to mention a few). An analysis of the German-into-Basque translation of multi-word expressions (Sanz-Villar 2024) observed that students’ dependence on MT systems was very high; i.e., traces of the MT proposals were found in the outputs of translation trainees. To specifically analyse the impact of MT proposals on students’ translations, two tasks were designed in 2021 in which nine students participated. Both tasks consisted of translating an excerpt of a children’s literary text from German into Basque. The first translation task was done with the aid of MT tools and the second was a from-scratch translation where the use of MT systems was not allowed. The goal of the present study is to compare the results of both tasks, focusing on the translation of phraseological units (PUs) to observe the features of students’ translations accomplished with and without the aid of MT tools. Since students used MT outputs during the first translation task, machine-translated texts are also compared to human-translated ones. The results show that the tendency to retain the phraseological character of the original text is stronger in professional translations than in machine-translated texts. In general, reliance on MT system outputs when translating PUs is quite strong among trainee translators. When comparing the results of the two tasks, it was observed that for some students, the number of techniques used to maintain the phraseological level of the original text increased in the translation done without MT. Interestingly, examples of creative translations were found in the translations done without the aid of MT tools.

A Report on E-C Translation of English Humorous Novels

This translation practice report selects a short humorous novel: This May Hurt a Little as the translation material and case selection scope. This translation practice report is divided into five chapters. The first chapter introduces the background and significance of the translation task. Chapter Two introduces the source text and the author. The third chapter introduces the translation process in detail. Before translation, the translator carefully studied the original text and determined the translation tools and translation theories, namely Newmark’s semantic translation and communicative translation. The fourth chapter, under the guidance of semantic translation and communicative translation, carries out specific case analysis from three levels : vocabulary, syntax and discourse. The last chapter summarizes the experience and gains of the translation practice.

Distilling Knowledge in Machine Translation of Agglutinative Languages with Backward and Morphological Decoders

Agglutinative languages often have morphologically complex words(MCWs) composed of multiple morphemes arranged in a hierarchical structure, posing significant challenges in translation tasks. We present a novel Knowledge Distillation approach tailored for improving the translation of such languages. Our method involves an encoder, a forward decoder, and two auxiliary decoders: a backward decoder and a morphological decoder. The forward decoder generates target morphemes autoregressively and is augmented by distilling knowledge from the auxiliary decoders. The backward decoder incorporates future context, while the morphological decoder integrates target-side morphological information. We have also designed a reliability estimation method to selectively distill only the reliable knowledge from these auxiliary decoders. Our approach relies on morphological word segmentation. We show that the word segmentation method based on unsupervised morphology learning outperforms the commonly used Byte Pair Encoding method on highly agglutinative languages in translation tasks. Our experiments conducted on English-Tamil, English-Manipuri, and English-Marathi datasets show that our proposed approach achieves significant improvements over strong Transformer-based NMT baselines.

Multi-Agent Approach to Political Discourse Translation: From Large Language Models to MAGIC-PTF System

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.

Baoulé Related Parallel Corpora for Machine Translation Tasks: Mtbci-1.0

According to the Ethnologue platform, we have 7,164 known living languages in the World, and not all of them have data available over the internet to facilitate Artificial Intelligence (AI) tasks such as Machine Translation (MT). Consequently, there is a need for thorough Data Engineering tasks for most of these languages. Especially, the Baoulé living language normalized as ISO 639-3 (bci) is not yet supported on popular worldwide free translation platform such as Microsoft Translator, nor on the Official Wikipedias. In this paper, we have proposed the "Baoulé Related Parallel Corpora for Machine Translation tasks: mtBCI-1.0" to make parallel Baoulé-related datasets available to the scientific community for AI tasks implying Machine Translation. We have shown that, after a brief presentation of the Baoulé language in the proposed approach, we will focus on the Data Engineering Process itself before providing a baseline proving that the collected data is of scientific interest.

The importance of coding and translation between programming languages in sequential activities of pre-service teachers: an approach

AbstractIn the last two decades, computational thinking has gained wide relevance in international educational systems. The inclusion of this new type of thinking poses educational challenges with some underlying research questions that need to be answered to meet these challenges with quality. Thus, this study focuses on analyzing the difficulties that teachers in initial training experience have, when carrying out translation tasks of programming languages used by certain educational robots, in this case, the Cubetto. For this purpose, a specific learning sequence has been designed to work with different programming languages (Cubetto, Bee-Bot, Scratch) and natural language. The work of early childhood and elementary trainee teachers in these tasks has been analyzed using a descriptive approach. The main results are: (1) some of the difficulties encountered are clearly caused by the Cubetto hardware (regardless of the language to which it is translated) and (2) the designed learning sequence has enabled coding skills to be improved remarkably. We conclude that translation tasks between programming languages are necessary in initial teacher training to improve their ability programming and their computational thinking, and for them to be able to detect the disadvantages and benefits of educational robots in their transposition to the classroom.

Challenges of Translating Idiomatic Expressions: A Cross-Linguistic Analysis at a University in Hanoi, Vietnam

This study investigates the challenges students face when translating idiomatic expressions between English and Vietnamese and explores how linguistic and cultural factors influence the translation process. Utilizing a combination of translation task analysis and focus group discussions, the research reveals that students frequently rely on literal translations, resulting in inaccurate or nonsensical outcomes due to their inability to grasp the figurative meanings of idioms. Cultural mismatches, particularly the absence of equivalent expressions in Vietnamese, further complicate translation efforts, leading to significant misunderstandings. Additionally, linguistic barriers, such as syntactic differences between English and Vietnamese, hinder students' ability to produce natural and accurate translations of idiomatic expressions. The findings emphasize the critical role of cultural knowledge and syntactic flexibility in effective idiomatic translation. To address these challenges, the study recommends that translation training programs place greater emphasis on idiomatic understanding and cross-cultural competence. This study contributes to the growing body of literature on idiomatic translation by highlighting the interplay of linguistic and cultural factors and suggesting strategies to enhance translation accuracy.

Traduire le théâtre de Gil Vicente, spécificités de traduction et possibles apports liés à l’intelligence artificielle

Translating the theater of Gil Vicente poses a significant challenge, that only specialists about the author, with some rare exceptions, have translated his plays. Between the nuances of the text, the archaisms of the language, and the specificities of theatrical writing, this playwright has seen few of his plays, also known as autos, translated into French. At a time when the relationship between the translator and technological advancements is becoming increasingly intertwined, it is important to analyze how certain software, particularly those of Artificial Intelligence, perform in the task of translation and what their impact might be on future translations of Gil Vicente.

Study on an Intelligent English Translation Method Using an Improved Convolutional Neural Network Model

This study uses an enhanced convolutional neural network (CNN) model for English translation. Traditional translation methods often struggle with complex language structures, prompting the adoption of deep learning techniques, particularly CNNs, in natural language processing. The research outlines CNN fundamentals and their relevance in language processing, elucidating the design and implementation of an improved CNN model. To address CNN limitations in maintaining coherence during the translation of lengthy texts, historical attention mechanisms are incorporated to enhance translation performance. Experimental validation conducted in MATLAB demonstrates notable improvements in translation task performance, evidenced by significant increases in BLEU scores. Results highlight the model's capacity to integrate contextual information, thereby enhancing translation coherence and accuracy. Additionally, the study establishes a mathematical framework for augmenting CNNs with attention mechanisms, providing valuable insights for the development of intelligent English translation systems.

Cross-Task Differences in Frontocentral Cortical Activations for Dynamic Balance in Neurotypical Adults.

Although significant progress has been made in understanding the cortical correlates underlying balance control, these studies focused on a single task, limiting the ability to generalize the findings. Different balance tasks may elicit cortical activations in the same regions but show different levels of activation because of distinct underlying mechanisms. In this study, twenty young, neurotypical adults were instructed to maintain standing balance while the standing support surface was either translated or rotated. The differences in cortical activations in the frontocentral region between these two widely used tasks were examined using electroencephalography (EEG). Additionally, the study investigated whether transcranial magnetic stimulation could modulate these cortical activations during the platform translation task. Higher delta and lower alpha relative power were found over the frontocentral region during the platform translation task when compared to the platform rotation task, suggesting greater engagement of attentional and sensory integration resources for the former. Continuous theta burst stimulation over the supplementary motor area significantly reduced delta activity in the frontocentral region but did not alter alpha activity during the platform translation task. The results provide a direct comparison of neural activations between two commonly used balance tasks and are expected to lay a strong foundation for designing neurointerventions for balance improvements with effects generalizable across multiple balance scenarios.

A Study on Benjamin’s Theory of Translation

Walter Benjamin is one of the most creative and important critics of the early 20th century, and his work “The Task of Translation” stands as one of the most influential translation theories. It has greatly influenced the traditional notions of faithfulness and the “original language first” principle, laying the theoretical foundation for deconstructionist translation theory.

Efficient Adaptation: Enhancing Multilingual Models for Low-Resource Language Translation

This study focuses on the neural machine translation task for the TR-EN language pair, which is considered a low-resource language pair. We investigated fine-tuning strategies for pre-trained language models. Specifically, we explored the effectiveness of parameter-efficient adapter methods for fine-tuning multilingual pre-trained language models. Various combinations of LoRA and bottleneck adapters were experimented with. The combination of LoRA and bottleneck adapters demonstrated superior performance compared to other methods. This combination required only 5% of the pre-trained language model to be fine-tuned. The proposed method enhances parameter efficiency and reduces computational costs. Compared to the full fine-tuning of the multilingual pre-trained language model, it showed only a 3% difference in the BLEU score. Thus, nearly the same performance was achieved at a significantly lower cost. Additionally, models using only bottleneck adapters performed worse despite having a higher parameter count. Although adding LoRA to pre-trained language models alone did not yield sufficient performance, the proposed method improved machine translation. The results obtained are promising, particularly for low-resource language pairs. The proposed method requires less memory and computational load while maintaining translation quality.

Unpaired data training enables super-resolution confocal microscopy from low-resolution acquisitions.

Supervised deep-learning models have enabled super-resolution imaging in several microscopic imaging modalities, increasing the spatial lateral bandwidth of the original input images beyond the diffraction limit. Despite their success, their practical application poses several challenges in terms of the amount of training data and its quality, requiring the experimental acquisition of large, paired databases to generate an accurate generalized model whose performance remains invariant to unseen data. Cycle-consistent generative adversarial networks (cycleGANs) are unsupervised models for image-to-image translation tasks that are trained on unpaired datasets. This paper introduces a cycleGAN framework specifically designed to increase the lateral resolution limit in confocal microscopy by training a cycleGAN model using low- and high-resolution unpaired confocal images of human glioblastoma cells. Training and testing performances of the cycleGAN model have been assessed by measuring specific metrics such as background standard deviation, peak-to-noise ratio, and a customized frequency content measure. Our cycleGAN model has been evaluated in terms of image fidelity and resolution improvement using a paired dataset, showing superior performance than other reported methods. This work highlights the efficacy and promise of cycleGAN models in tackling super-resolution microscopic imaging without paired training, paving the path for turning home-built low-resolution microscopic systems into low-cost super-resolution instruments by means of unsupervised deep learning.

Translating Traditional Folk Poetry: The Perception of Translators and Translation Students and the Reality in Saudi Society

Despite the massive presence and influence of traditional folk poetry (TFP) within Saudi society, the efforts to translate such works remain very limited. While many authors and researchers have examined the impact of this literary genre and illustrated its prominence, academic investigations into the difficulties of translating TFP works are almost nonexistent. This is also true about studies that focus on the overarching sentiment regarding these efforts among those who are associated with the field of language and translation and among members of the general public. This paper thus aims at examining these perspectives and outlining the challenges translation trainees face when performing tasks in this field. The data for this study was collected through the dissemination of a questionnaire and the analysis of a translation task assigned to university students in an academic program for English language and translation. The qualitative and quantitative analysis of the data shows that the level of acceptance of TFP translation efforts is strongly connected to the degree of association with the field of languages and translation, with individuals from outside these circles being less supportive of initiatives in this domain. Additionally, the paper identified several obstacles and challenges encountered by translation students working on TFP texts. The paper concludes with a number of recommendations and suggestions that can, hopefully, contribute to the efforts to raise awareness about translating TFP and to the process of improving the current state of translation within this field.