Synthetic Aperture Radar (SAR) imagery provides all-weather, all-day, and high-resolution imaging capabilities but its unique imaging mechanism makes interpretation heavily reliant on expert knowledge, limiting interpretability, especially in complex target tasks. Translating SAR images into optical images is a promising solution to enhance interpretation and support downstream tasks. Most existing research focuses on scene-level translation, with limited work on object-level translation due to the scarcity of paired data and the challenge of accurately preserving contour and texture details. To address these issues, this study proposes a keypoint-guided diffusion model (KeypointDiff) for SAR-to-optical image translation of unpaired aircraft targets. This framework introduces supervision on target class and azimuth angle via keypoints, along with a training strategy for unpaired data. Based on the classifier-free guidance diffusion architecture, a class-angle guidance module (CAGM) is designed to integrate class and angle information into the diffusion generation process. Furthermore, adversarial loss and consistency loss are employed to improve image fidelity and detail quality, tailored for aircraft targets. During sampling, aided by a pre-trained keypoint detector, the model eliminates the requirement for manually labeled class and azimuth information, enabling automated SAR-to-optical translation. Experimental results demonstrate that the proposed method outperforms existing approaches across multiple metrics, providing an efficient and effective solution for object-level SAR-to-optical translation and downstream tasks. Moreover, the method exhibits strong zero-shot generalization to untrained aircraft types with the assistance of the keypoint detector.
Increasingly, more and more people are turning to large language models (LLMs) for healthcare advice and consultation, making it important to gauge the efficacy and accuracy of the responses of LLMs to such queries. While there are pre-existing medical benchmarks literature which seeks to accomplish this very task, these benchmarks are almost universally in English, which has led to a notable gap in existing literature pertaining to multilingual LLM evaluation. Within this work, we seek to aid in addressing this gap with Cancer-Myth-Indic, an Indic language benchmark built by translating a 500-item subset of Cancer-Myth, sampled evenly across its original categories, into five under-served but widely used languages from the subcontinent (500 per language; 2,500 translated items total). Native-speaker translators followed a style guide for preserving implicit presuppositions in translation; items feature false presuppositions relating to cancer. We evaluate several popular LLMs under this presupposition stress.
Despite the remarkable progress of modern machine translation (MT) systems on general-domain texts, translating structured LaTeX-formatted documents remains a significant challenge. These documents typically interleave natural language with domain-specific syntax, such as mathematical equations, tables, figures, and cross-references, all of which must be accurately preserved to maintain semantic integrity and compilability. In this paper, we introduce LaTeXTrans, a collaborative multi-agent system designed to address this challenge. LaTeXTrans ensures format preservation, structural fidelity, and terminology consistency through six specialized agents: 1) a Parser that decomposes LaTeX into translation-friendly units via placeholder substitution and syntax filtering; 2) a Translator, Validator, Summarizer, and Terminology Extractor that work collaboratively to ensure context-aware, self-correcting, and terminology-consistent translations; 3) a Generator that reconstructs the translated content into well-structured LaTeX documents. Experimental results show that LaTeXTrans outperforms mainstream MT systems in both translation accuracy and structural preservation. The source code, the online demonstration platform, and a demo video are publicly available.
Abstract The aim of this article is to show how sound can translate knowledge, not with words but sensorially, with the whole body. Together with the new ways of understanding translation not only through words but with the whole body and senses, the article bases its argument on the new approaches to sound and music of the so-called “Sound studies” and “Sonic art” and applies the notion of “atmosphere” to Cecilia Vicuña’s artworks. She uses her voice and the voices of others as a means to translate without words and thus creates atmospheres that are affective translations of the land and the environment, of the human and the non-human. Her atmospheres are spaces in which she translates corporeally.
This article explores the different ways George Orwell’s works have entered Ukrainian culture, with a particular focus on the translations that have introduced his sharp political commentary to Ukrainian readers. Apart from translations, we will examine other channels of interpretation, such as various adaptations and critical analyses, which have further expanded Orwell’s reach and influence in Ukraine. By analyzing these various forms of transmission, we would like to shed light on how Orwell’s themes have resonated in a Ukrainian context, and how his works continue to inspire reflection and discourse in a nation with its own complex history of sociopolitical struggle. This study represents the first attempt to gather and analyze the various interpretations of Orwell’s legacy within Ukraine.
Neural Machine Translation (NMT) is the task of translating a text from one language to another with the use of a trained neural network. Several existing works aim at incorporating external information into NMT models to improve or control predicted translations (e.g. sentiment, politeness, gender). In this work, we propose to improve translation quality by adding another external source of information: the automatically recognized emotion in the voice. This work is motivated by the assumption that each emotion is associated with a specific lexicon that can overlap between emotions. Our proposed method follows a two-stage procedure. At first, we select a state-of-the-art Speech Emotion Recognition (SER) model to predict dimensional emotion values from all input audio in the dataset. Then, we use these predicted emotions as source tokens added at the beginning of input texts to train our NMT model. We show that integrating emotion information, especially arousal, into NMT systems leads to better translations.
Multilingual large language models (LLMs) are great translators, but this is largely limited to high-resource languages. For many LLMs, translating in and out of low-resource languages remains a challenging task. To maximize data efficiency in this low-resource setting, we introduce Mufu, which includes a selection of automatically generated multilingual candidates and an instruction to correct inaccurate translations in the prompt. Mufu prompts turn a translation task into a postediting one, and seek to harness the LLM's reasoning capability with auxiliary translation candidates, from which the model is required to assess the input quality, align the semantics cross-lingually, copy from relevant inputs and override instances that are incorrect. Our experiments on En-XX translations over the Flores-200 dataset show LLMs finetuned against Mufu-style prompts are robust to poor quality auxiliary translation candidates, achieving performance superior to NLLB 1.3B distilled model in 64% of low- and very-low-resource language pairs. We then distill these models to reduce inference cost, while maintaining on average 3.1 chrF improvement over finetune-only baseline in low-resource translations.
In this paper, we present Cross Language Agent -- Simultaneous Interpretation, CLASI, a high-quality and human-like Simultaneous Speech Translation (SiST) System. Inspired by professional human interpreters, we utilize a novel data-driven read-write strategy to balance the translation quality and latency. To address the challenge of translating in-domain terminologies, CLASI employs a multi-modal retrieving module to obtain relevant information to augment the translation. Supported by LLMs, our approach can generate error-tolerated translation by considering the input audio, historical context, and retrieved information. Experimental results show that our system outperforms other systems by significant margins. Aligned with professional human interpreters, we evaluate CLASI with a better human evaluation metric, valid information proportion (VIP), which measures the amount of information that can be successfully conveyed to the listeners. In the real-world scenarios, where the speeches are often disfluent, informal, and unclear, CLASI achieves VIP of 81.3% and 78.0% for Chinese-to-English and English-to-Chinese translation directions, respectively. In contrast, state-of-the-art commercial or open-source systems only achieve 35.4% and 41.6%. On the extremely hard dataset, where other systems achieve under 13% VIP, CLASI can still achieve 70% VIP.
As a type of figurative language, an East Asian idiom condenses rich cultural background into only a few characters. Translating such idioms is challenging for human translators, who often resort to choosing a context-aware translation from an existing list of candidates. However, compiling a dictionary of candidate translations demands much time and creativity even for expert translators. To alleviate such burden, we evaluate if GPT-4 can help generate high-quality translations. Based on automatic evaluations of faithfulness and creativity, we first identify Pareto-optimal prompting strategies that can outperform translation engines from Google and DeepL. Then, at a low cost, our context-aware translations can achieve far more high-quality translations per idiom than the human baseline. We open-source all code and data to facilitate further research.
On Wikipedia, articles are categorized to aid readers in navigating content efficiently. The manual creation of new categories can be laborious and time-intensive. To tackle this issue, we built language models to translate Wikipedia categories from English to Vietnamese with a dataset containing 15,000 English-Vietnamese category pairs. Subsequently, small to medium-scale Transformer pre-trained models with a sequence-to-sequence architecture were fine-tuned for category translation. The experiments revealed that OPUS-MT-en-vi surpassed other models, attaining the highest performance with a BLEU score of 0.73, despite its smaller model storage. We expect our paper to be an alternative solution for translation tasks with limited computer resources.
<p>This research work seeks to ascertain the translatability of living together in La promesse de Malingo by Pierre Fandio into English as an asset for the sustainability of the united Cameroon. Five elements helped the researcher attain the above objective: multilingualism, implicit, explicit, bilingualism and onomastics. This study poses the problem of the strategies used by a translator to produce an equivalent effect in the minds of the receptors of the target text. Two hypotheses carried the researcher through the present research: the first one stated that Bilingualism, multilingualism, explicit, onomastics and implicit are situations of living together that exist in the novel La Promesse de Malingo. The second said that macro and micro translation strategies could be relevant for the translation of togetherness into English in the novel. A micro textual analysis of the novel with 50 excerpts obtained from the novel confirmed the presence of the above elements as situations of living together found in the novel. Foreignisation as a macro translation strategy had an 86% usage while literal translation as a micro translation strategy carried a 50% occurrence which is the highest as compared to other micro translation strategies. The polysystem theory, with a 60% usage, happens to be the key translation theory of this work, showing that although words are important in translation, the context plays a vital role. The latter conclusion helps ascertain the translatability of living together and thus validates the hypotheses of this research. Recommendations and suggestions for further research were thus made. </p><p>La présente étude intitulée : vers une traduction du vivre ensemble au sein de la crise identitaire s’inscrit dans le domaine de la traduction littéraire. Elle vise à déterminer la traduisibilité du vivre ensemble dans l’œuvre La Promesse de Malingo de Pierre Fandio vers l’Anglais, comme un atout pour le maintien du Cameroun uni. Cinq éléments nous ont permis d'atteindre cet objectif : le multilinguisme, l'implicite, l'explicite, le bilinguisme et l'onomastique. Cette étude pose le problème des stratégies utilisées par un traducteur pour produire un effet équivalent dans l'esprit des récepteurs du texte cible. Deux hypothèses ont guidé le chercheur tout au long de la présente recherche : la première affirme que le bilinguisme, le multilinguisme, l'explicite, l'onomastique et l'implicite sont des situations de vivre ensemble qui existent dans le roman La Promesse de Malingo. La seconde affirme que les stratégies de macro et micro-traduction pourraient être pertinentes pour la traduction du vivre ensemble en anglais dans le roman. Une analyse micro textuelle constitué de 50 extraits du roman a confirmé la présence des éléments susmentionnés comme situations de vivre ensemble dans le roman. L'étrangéisation en tant que macro-stratégie de traduction a été utilisée à 86 %. Cependant, la traduction littérale comme micro-stratégie de traduction a été utilisée à 50 %, ce qui est le taux le plus élevé par rapport à d'autres micro-stratégies de traduction. La théorie du polysystème, utilisée à 60 %, s'est avérée être la théorie de traduction phare de ce travail ; ce pourcentage démontre bien que les mots sont tout aussi important que le contexte en traduction. Cette conclusion permet de vérifier la traduisibilité du vivre ensemble et donc de valider les hypothèses de cette recherche. Des recommandations et des suggestions pour la suite de la recherche ont ainsi été formulées. </p><p> </p><p><strong> Article visualizations:</strong></p><p><img src="/-counters-/soc/0722/a.php" alt="Hit counter" /></p>
The ability to stop already-initiated actions is a key cognitive control ability. Recent work on human action-stopping has been dominated by two controversial debates. First, the contributions (and neural signatures) of attentional orienting and motor inhibition after stop-signals are near-impossible to disentangle. Second, the timing of purportedly inhibitory (neuro)physiological activity after stop-signals has called into question which neural signatures reflect processes that actually contribute to action-stopping. Here, we propose that a two-stage model of action-stopping - proposed by Schmidt and Berke (2017) based on subcortical rodent recordings - may resolve these controversies. Translating this model to humans, we first argue that attentional orienting and motor inhibition are inseparable because orienting to salient events like stop-signals automatically invokes broad motor inhibition, reflecting a fast-acting, ubiquitous Pause process. We then argue that inhibitory signatures after stop-signals differ in latency because they map onto two sequential stages: the salience-related Pause and a slower, stop-specific Cancel process. We formulate the model, discuss recent supporting evidence in humans, and interpret existing data within its context.
Ribosomes are complex ribozymes that interpret genetic information by translating messenger RNA (mRNA) into proteins. Natural variation in ribosome composition has been documented in several organisms and can arise from several different sources. A key question is whether specific control over ribosome heterogeneity represents a mechanism by which translation can be regulated. We used RiboMeth-seq to demonstrate that differential 2′-O-methylation of ribosomal RNA (rRNA) represents a considerable source of ribosome heterogeneity in human cells, and that modification levels at distinct sites can change dynamically in response to upstream signaling pathways, such as MYC oncogene expression. Ablation of one prominent methylation resulted in altered translation of select mRNAs and corresponding changes in cellular phenotypes. Thus, differential rRNA 2′-O-methylation can give rise to ribosomes with specialized function. This suggests a broader mechanism where the specific regulation of rRNA modification patterns fine tunes translation. Dynamic changes in 2′-O-methylation of rRNA in human cells lead to ribosome heterogeneity and result in altered translation of select mRNAs, correlating with changes in cellular phenotypes.
This introduction to the 10th issue of Linguistica Antverpiensia New Series – Themes in translation Studies (LANS-TTS) begins by discussing the central concept of community translation, highlighting its terminological ambiguity. This is in part due to the already well-established field of community interpreting where the term is often used to mean the written translation of public information for immigrants. It is also an indication of the terminological instability typical of an emerging paradigm. For example, community translation is used more or less synonymously with such terms as translation crowdsourcing, user-generated translation and collaborative translation. The meaning of the term as we discuss in this issue can be best specified when the concept is anchored in the context of Web 2.0 (second generation web-technologies). This in turn acknowledges its intrinsic tie to online communities and directs us to new dynamics resulting from general Internet users acting as translators. While participants in community translation are not necessarily all unpaid, untrained volunteers community translation is used by some organisations as a mechanism to obtain free translations by going outside the professional translation sphere. To this end the ethical question of profit-making enterprises accessing free labour on the pretext of openness and sharing remains. That said, the author believes community translation is far more than a dilettante, anti-professional movement. Building on the emerging picture from the contributions in this volume, the author suggests some of the future directions that research on community translation might take, emphasising the need to reflect on the current translation practices and be open to the new developments and opportunities arising from the free and social Internet.
This paper deals with the translation of the English past progressive into Arabic
by examining both theoretical and textual considerations. First, it shows how the English aspect formally corresponds to the auxiliary verb كا ن ‘was’ plus the simple present (SP) form or the active participle (AP) form. Second, it is argued that the choice between the SP and the AP is subject to several grammatical and semantic constraints on Arabic verbs: [+/- transitive], [+/- telic], [+/- completed], and [+/- manner of motion]. Third, the textual data (70 examples) drawn from two existing Arabic translations of Leonardo DaVinci by Walter Isaacs (2017) and Hard Choices by Hillary Clinton (2014) indicates that several translation procedures are employed to render the English past progressive, mainly including the past simple (48.57%), present simple (22.85%), formal correspondence (18.57%), and lexicalizing (7.14%). Finally, the qualitative analysis reveals that the progressiveness, emphasis, and dramatization that the English past progressive aspect may communicate are seriously compromised in Arabic translation. While there may be cases where some mismatches between English and Arabic verbs exist in terms of progressiveness which may call for the use of past simple or lexicalizing, the formal correspondence procedure is claimed to be the most valid and appropriate for capturing the functions of the English past progressive.
Among many definitions, translation can be described as decision making, which involves the concepts of problem solving, strategies, and choices, situating translation as a process which is oriented to study what goes on in the mind of the translators. Then, decision making can be addressed in studies concerning the translator, rather than the product, and can be tackled within the field of translation process research. This is already the case for think-aloud protocols (TAP), studied by many scholars within the framework of decision making. In spite of the criticism it gets, it is obvious that TAP provide rich data on decision making in translation, enhancing a wider perspective on the process-oriented approaches. Based on this perspective, then, TAP can also be studied within the situated and distributed cognition approaches to translation as a valuable research method that has access to the “black box” that will also provide an awareness of the fact that translation decision making and problem solving are not only restricted to texts. In this review, how decision making and translators are studied in translation process research will be reviewed, making some suggestions for future studies.
Sudhansu Bala Das, Divyajoti Panda, Tapas Kumar Mishra
et al.
Machine Translation (MT) system generally aims at automatic representation of source language into target language retaining the originality of context using various Natural Language Processing (NLP) techniques. Among various NLP methods, Statistical Machine Translation(SMT). SMT uses probabilistic and statistical techniques to analyze information and conversion. This paper canvasses about the development of bilingual SMT models for translating English to fifteen low-resource Indian Languages (ILs) and vice versa. At the outset, all 15 languages are briefed with a short description related to our experimental need. Further, a detailed analysis of Samanantar and OPUS dataset for model building, along with standard benchmark dataset (Flores-200) for fine-tuning and testing, is done as a part of our experiment. Different preprocessing approaches are proposed in this paper to handle the noise of the dataset. To create the system, MOSES open-source SMT toolkit is explored. Distance reordering is utilized with the aim to understand the rules of grammar and context-dependent adjustments through a phrase reordering categorization framework. In our experiment, the quality of the translation is evaluated using standard metrics such as BLEU, METEOR, and RIBES
As the use of deep learning techniques has grown across various fields over the past decade, complaints about the opaqueness of the black-box models have increased, resulting in an increased focus on transparency in deep learning models. This work investigates various methods to improve the interpretability of deep neural networks for Natural Language Processing (NLP) tasks, including machine translation and sentiment analysis. We provide a comprehensive discussion on the definition of the term interpretability and its various aspects at the beginning of this work. The methods collected and summarised in this survey are only associated with local interpretation and are specifically divided into three categories: (1) interpreting the model’s predictions through related input features; (2) interpreting through natural language explanation; (3) probing the hidden states of models and word representations.
The purpose of late phase clinical trials is to generate evidence of sufficient validity and generalisability to be translated into practice and policy to improve health outcomes. It is therefore crucial that the chosen endpoints are meaningful to the clinicians, patients and policymakers that are the end-users of evidence generated by these trials. The choice of endpoints may be improved by understanding their characteristics and properties. This narrative review describes the evolution, range and relative strengths and weaknesses of endpoints used in late phase trials. It is intended to serve as a reference to assist those designing trials when choosing primary endpoint(s), and for the end-users charged with interpreting these trials to inform practice and policy.