Reliable Translation Research Articles

Analysis of Two Translation Applications : Why is DeepL Translate more accurate than Google Translate?

DeepL Translate and Google Translate are two leading machine translation tools. The focus of this research is to analyze the accuracy of translation results provided by DeepL Translation and Google Translation specifically in translating English to Indonesian. This research used a qualitative approach of document analysis and interviews. The advanced neural machine translation technology of DeepL, by utilizing extensive data, enables it to recognize language nuances and provide contextually accurate translations. In contrast, Google Translate, despite having grown to be supported by hundreds of languages, often struggles with complex sentences and idiomatic expressions. DeepL's accuracy and natural-sounding translations make it a top choice for professional and detailed translations. The study concludes that DeepL's focus on quality and accuracy, rather than the breadth of language support, makes it a more reliable translation tool.

Research on English Translation Optimization Algorithm Based on Statistical Machine Learning

In the study titled Research on English Translation Optimization Algorithm Based on Statistical Machine Learning: IAAM-NN (Integrating Advanced Attention Mechanisms with Neural Networks), we explore the fusion of advanced attention mechanisms with neural networks to enhance English translation accuracy. This research delves into the intersection of statistical machine learning and language processing, presenting a novel approach termed IAAM-NN. This method capitalizes on the strengths of neural networks in learning complex patterns and the refined attention mechanisms’ ability to accurately map contextual relationships within text. The core objective is to address the challenges faced in traditional translation algorithms – primarily context misinterpretation and semantic inaccuracies. By harnessing the power of advanced attention mechanisms, the IAAM-NN algorithm effectively deciphers nuanced linguistic structures, ensuring more accurate and contextually relevant translation outputs. This study demonstrates the potential of combining neural network models with enhanced attention processes, illustrating significant improvements in translation quality compared to standard machine learning approaches. The implementation of IAAM-NN marks a step forward in the realm of machine translation, offering insights into developing more sophisticated and reliable translation tools in the future.

The neural network algorithm-based quality assessment method for university English translation

ABSTRACT These results highlight the transformative potential of neural network algorithms in providing consistency and transparency while reducing the inherent subjectivity in human evaluations, revolutionizing translation quality assessment in academia. The findings have significant implications for academia, as reliable translation quality evaluations are crucial for fostering cross-cultural knowledge exchange. However, challenges such as domain-specific adaptation require further investigation to improve and maximize the effectiveness of this novel approach, ultimately enhancing the accessibility of academic content and promoting global academic discourse. The proposed method involves using neural network algorithms for assessing college-level English translation quality, starting with data collection and preparation, developing a neural network model, and evaluating its performance using human assessment as a benchmark. The study employed both human evaluators and a neural network model to assess the quality of translated academic papers, revealing a strong correlation (0.84) between human and model assessments. These findings suggest the model’s potential to enhance translation quality in academic settings, though additional research is needed to address certain limitations. The results show that the Neural Network-Based Model achieved higher scores in accuracy, precision, F-measure, and recall compared to Traditional Manual Evaluation and Partial Automated Model, indicating its superior performance in evaluating translation quality.

Reliable multi-modal medical image-to-image translation independent of pixel-wise aligned data.

The current mainstream multi-modal medical image-to-image translation methods face a contradiction. Supervised methods with outstanding performance rely on pixel-wise aligned training data to constrain the model optimization. However, obtaining pixel-wise aligned multi-modal medical image datasets is challenging. Unsupervised methods can be trained without paired data, but their reliability cannot be guaranteed. At present, there is no ideal multi-modal medical image-to-image translation method that can generate reliable translation results without the need for pixel-wise aligned data. This work aims to develop a novel medical image-to-image translation model that is independent of pixel-wise aligned data (MITIA), enabling reliable multi-modal medical image-to-image translation under the condition of misaligned trainingdata. The proposed MITIA model utilizes a prior extraction network composed of a multi-modal medical image registration module and a multi-modal misalignment error detection module to extract pixel-level prior information from training data with misalignment errors to the largest extent. The extracted prior information is then used to construct a regularization term to constrain the optimization of the unsupervised cycle-consistent Generative Adversarial Network model, restricting its solution space and thereby improving the performance and reliability of the generator. We trained the MITIA model using six datasets containing different misalignment errors and two well-aligned datasets. Subsequently, we conducted quantitative analysis using peak signal-to-noise ratio and structural similarity as metrics. Moreover, we compared the proposed method with six other state-of-the-art image-to-image translationmethods. The results of both quantitative analysis and qualitative visual inspection indicate that MITIA achieves superior performance compared to the competing state-of-the-art methods, both on misaligned data and aligned data. Furthermore, MITIA shows more stability in the presence of misalignment errors in the training data, regardless of their severity or type. The proposed method achieves outstanding performance in multi-modal medical image-to-image translation tasks without aligned training data. Due to the difficulty in obtaining pixel-wise aligned data for medical image translation tasks, MITIA is expected to generate significant application value in this scenario compared to existingmethods.

A porous elastomer with a cavity array for three-dimensional plantar force sensing

Human gait is an important indicator for diagnosis of various movement-related disorders. Three-dimensional (3D) plantar force monitoring enables real-time recording of contact forces in both the normal and tangential directions in each step, providing useful clues for identification of irregular gait patterns or gait-related health issues. Most flexible sensors for gait monitoring focus on contact forces in the normal direction, being limited to detect forces in the tangential direction. Herein we have developed a porous elastomer with a cavity array (PECA) to measure 3D contact forces for gait analysis. The sensor consists of two fabric electrodes and a dielectric layer of porous Ecoflex elastomer containing a 5 × 5 cavity array, forming four parallel-plate capacitor units in a soft construction. Therefore, a 3D force in any arbitrary direction generates four unique capacitive electrical signals. Importantly, the PECA includes a well-designed combination of micrometer-scale pores and millimeter-sized cavities, dramatically improving the detection sensitivity and linear elasticity range. We have demonstrated high sensitivities of 0.41 N−1 and 0.23 N−1 in the normal and tangential directions, respectively. The flexible PECA-based sensors have been tested to sensitively distinguish various standing postures and different stride lengths for gait analysis, promising reliable translation into areas such as healthcare monitoring, sports training, and rehabilitation engineering.

Evaluation of the Turkish soundscape circumplex model, and the effects of sound source dominance, sequence, and gender on the perceived affective quality

The International Standard Organization (ISO) standards series onsoundscapefor the identification, data collection and data analysis present a general framework for the soundscape studies that are conducted globally. Perceived Affective Qualities (PAQs) including 8 soundscape perceptual attributes in ISO/TS 12913–2:2018 and ISO/TS 12913–3:2019 are published in English language. Therefore, the ‘Soundscape Attributes Translation Project’ focuses on the reliable standardized translations of the PAQs in other languages. In the previously published Turkish translation and validation study, the PAQs were presented as; ‘hareketli’ (eventful), ‘coşkulu’ (vibrant), ‘keyifli’ (pleasant), ‘sakin’ (calm), ‘durağan’ (uneventful), ‘tekdüze’ (monotonous), ‘rahatsız edici’ (annoying), and ‘karmaşık’ (chaotic). Within the scope of this follow-up study, firstly, sound sources in each audio stimulus are identified according to their dominance to explore their relation with the Soundscape Circumplex Model (SCM) quadrants and found that the human dominant audio stimuli are evaluated as ‘vibrant’ in SCM. Other noise dominating audio stimuli are categorized as ‘monotonous’ while traffic combined with other noise audio stimuli are evaluated as ‘chaotic’. When the audio stimuli evaluated as ‘calm’ are investigated, they are either purely dominated by natural sounds or mixture of each category dominated by natural sounds. Secondly, a comparison between the Turkish and the English participants’ evaluation of audio stimuli according to the SCM is done and found that they are similar except two audio stimuli. Furthermore, the effect of audio stimulus sequence and gender on 8 PAQs are statistically analysed through Kruskall-Wallis and Mann-Whitney tests; respectively and found that both are significant. The sequences shifting from ‘chaotic’ to ‘calm’ or vice versa frequently are seen to be evaluated with higher scores whereas more homogenous ones in sequence, i.e. audio stimuli having the same dominating sound source are sequenced together mostly are graded with lower scores of 8 PAQs except ‘chaotic’ compared to drastic changes in the sound source and found that men grade ‘coşkulu’ (vibrant), ‘durağan’ (uneventful), ‘tekdüze’ (monotonous), ‘rahatsız edici’ (annoying) and ‘karmaşık’ (chaotic) more than women by 5, 11, 24, 8 and 10 points, respectively according to the median values. Finally, conformance to the SCM is evaluated by the principal component analysis (PCA) and found that it can be reduced to two dimensions named as “pleasantness” and “eventfulness” in expected order but the main PAQs pleasant-annoying and eventful-uneventful are not aligned indicating the lack of achieving the ideal circumplexity. It is believed that the results of this study could contribute to both national and international future soundscape research with a focus on listening tests applying PAQs.

Evaluating optimal reference translations

Abstract The overall translation quality reached by current machine translation (MT) systems for high-resourced language pairs is remarkably good. Standard methods of evaluation are not suitable nor intended to uncover the many translation errors and quality deficiencies that still persist. Furthermore, the quality of standard reference translations is commonly questioned and comparable quality levels have been reached by MT alone in several language pairs. Navigating further research in these high-resource settings is thus difficult. In this paper, we propose a methodology for creating more reliable document-level human reference translations, called “optimal reference translations,” with the simple aim to raise the bar of what should be deemed “human translation quality.” We evaluate the obtained document-level optimal reference translations in comparison with “standard” ones, confirming a significant quality increase and also documenting the relationship between evaluation and translation editing.

The influence of charge on the translation of the sandwich ELISA approach to electronic biosensors

The sandwich approach, whereby an antigen is captured by a primary antibody and detected by a secondary antibody, is commonly used to improve the selectivity and sensitivity of enzyme-linked immunosorbent assays (ELISA). This work details the experimental factors that impact the reliable translation of this sandwich approach to two commonly used electronic biosensors, namely potentiometric and impedimetric biosensors. Previous studies have demonstrated the Debye screening limitations associated with potentiometric biosensors. However, the correlation between the ionic strength of the measurement buffer and the impedimetric biosensing response has not been studied. Potentiometric biosensors were able to successfully detect the primary antibody and the target antigen by decreasing the ionic strength of the phosphate buffered saline (PBS) measurement buffer from 1x PBS to 0.01x PBS. However, the secondary antibody used for the selective signal amplification was not reliably detected. Therefore, the sandwich approach is not viable for potentiometric sensing at biologically relevant ionic strengths, due to the Debye screening effect. Alternatively, decreasing the ionic strength of the measurement buffer allowed for the successful translation of the sandwich approach to impedimetric biosensors. Impedimetric biosensing literature typically attributes a measured increase in the charge transfer resistance to an increase in the thickness of the immobilized biolayer. However, this work highlights the influence that both the charge and thickness of the biolayer have on the transport of the redox couple. Decreasing the ionic strength of the measurement buffer lowers the molecular charge screening effect. This permits the transport of a positively charged redox probe through a negatively charged immobilized biolayer via migration and diffusion. The results demonstrate that the use of a buffer at a lower, yet biologically relevant ionic strength allows for the successful translation of the sandwich approach to impedimetric biosensors.

Translation and validation of the Hungarian version of the Birth Satisfaction Scale-Revised (BSS-R)

BackgroundEvidence relating maternal birth experience to a range of maternal and neonatal outcomes is increasingly compelling. Consequently valid and reliable self-report of birth experience from the mothers perspective is critical. AimThe current study sought to translate and validate a Hungarian-language version of the Birth Satisfaction Scale-Revised (BSS-R). MethodFollowing forward and backwards translation into Hungarian, the Hungarian BSS-R (HU-BSS-R) was administered to women in a major Transylvanian hospital maternity unit within 72 h postpartum. Key psychometric characteristics were then examined in relation to factor structure, divergent and convergent validity, internal consistency, and known-groups discriminant validity. ResultsTwo-hundred and thirty-two women completed the HU-BSS-R. Confirmatory factor analysis revealed the HU-BSS-R to offer an excellent fit to data for the established tri-dimensional measurement model. The HU-BSS-R was also found to offer excellent convergent and divergent validity and known-groups discriminant validity. No significant differences were observed between internal consistency observations between the current study and the original UK validation study. ConclusionsThe HU-BSS-R is a valid and reliable translation of the original BSS-R, it has proved itself to have excellent psychometric properties and is suitable for use in the Hungarian maternity context.

Enhancing quality of life and medication adherence for people living with HIV: the impact of an information system

BackgroundThe widespread availability of antiretroviral therapy has led to improvements in life expectancy and thus an increase in the number of people living with HIV/AIDS (PLWHA) worldwide. However, a similar increase in the number of newly-diagnosed patients in Cyprus suggests the need for solutions designed to improve monitoring, planning, and patient communication. In this study, we aimed to determine whether the use of an information system to manage PLWHA might contribute to improved quality of life and critical adherence to prescribed drug regimens and ongoing medical care.MethodsA randomized controlled trial study was conducted in Cyprus based on information that we collected using the highly valid and reliable Greek translation of the World Health Organization (WHO) Quality of Life (QOL) HIV-BREF questionnaire to assess sociodemographic variables and patient compliance. We distributed 200 questionnaires before implementing a Health Medical Care (HMC) information system at our clinic. Six months after implementing this system, 68 of the completed questionnaires were selected, including two groups of 34 participants who had been assigned at random to the intervention or the control group. Participants included PLWHA aged ≥ 18 years who had been receiving antiretroviral therapy for more than 12 months between July 15, 2020, and July 15, 2022.ResultsThe changes in baseline to six-month scores reported for the intervention group were significantly higher than in the control group in all six subscales assessed with the WHOQOL-HIV-BREF questionnaire, as well as in the assessment of compliance. Furthermore, compliance with treatment was associated with higher scores in the questionnaire subscales, including physical health, psychological health, degree of autonomy, social relationships, life circumstances, and spirituality/religious/personal beliefs. We also identified specific demographic factors and behaviors that were associated with better compliance with scheduled medical care and the prescribed drug regimen. Specifically, men exhibited better compliance than women and younger PLWHA exhibited better compliance than the elderly as did individuals who reported a higher level of educational attainment. Additionally, individuals who did not use addictive substances, consumed less alcohol, and were managed using the monitoring information system all exhibited better compliance compared to those in the control group.ConclusionThe results of this study suggest that management of PLWHA via the use of an information system can contribute to improved QOL and drug compliance.

DOCUMENT TRANSLATION FOR CHARITABLE ORGANIZATIONS ASSISTING UKRAINIAN REFUGEES: CHALLENGES AND STRATEGIES

The Russian-Ukrainian war displaced millions of Ukrainians many of whom now rely on international and local organizations in their host countries for help with daily problems of integration. One such aspect of charitable work is providing translation services. This article analyses the challenges and strategies of English, Ukrainian, and Russian document translation for charitable organizations assisting Ukrainian refugees. Drawing on professional experience of collaboration with a charitable organization, the author discusses communication that ensures efficient management of translation projects, outlines the certification procedure that confirms the translator’s responsibility for the accuracy of translation, and examines the types of documents that Ukrainian refugees commonly need to have translated. The author describes in detail the formatting of the translated documents that is required to make them accessible and easily comprehensible. The article delves into the challenges that the translator is faced with when working with the documents of Ukrainian refugees, focusing on such issues as illegible handwriting and stamps, the low quality of the images of the original documents, typographical errors, abbreviations and terms, formatting intricacies, temporal limitations, and confidentiality. Analysing the process of proofreading that improves the quality of translation, the author argues that translation is a social activity that requires collaboration between the translator, fellow professionals, legal experts, clients, local actors and stakeholders. Finally, this article explores the tools and strategies that enable translators to work consistently and efficiently. It examines how the translator can use machine translators, online corpus dictionaries, document translation templates, guidelines, and other aids to provide reliable, timely, accurate, and helpful translation services to Ukrainian refugees.

Vaivorykštė be saulės: tekstinė refrakcija sovietmečio vertimuose

Light waves change direction—refract—as they pass through different environments. Similarly, Western literary works, even after leaking through the barrier of the Soviet Index Librorum Prohibitorum, never reached the USSR reader in their authentic form. The most evident components of the unfavourable environment were the proxy of the Moscow Raduga and Progress publishing houses, ideologically committed literary criticism, the hunt for honorariums, the lack of reference books, and the poverty of printing technology. All these adverse factors together caused the chronic syndrome of texts that seemed authentic but in fact were not. So far, individual agents and effects of textual refraction have been researched, and interesting case studies have been carried out. However, there is a lack of both a systematic classification of the different agents of influence and an assessment of their long-term effects. The methodological study of refraction-affected editions tests the possibility of integrating the concept of textual condition with the assertions of the postcolonial theory regarding secondary cultural colonisation. The paper attempts to raise the hypothesis that the efforts of professional translators and editors to perform their work with the utmost craftsmanship had sometimes paradoxically reinforced the refraction of the overall meaning of major literary works. That resulted from the contrast between a relatively reliable translation and the implications generated by the bibliographic code and metatexts, as well as from the canonisation of translations that had become iconic in the Soviet-era.

Applying automated machine translation to educational video courses

We studied the capability of automated machine translation in the online video education space by automatically translating Khan Academy videos with state-of-the-art translation models and applying text-to-speech synthesis and audio/video synchronization to build engaging videos in target languages. We also analyzed and established two reliable translation confidence estimators based on round-trip translations in order to efficiently manage translation quality and reduce human translation effort. Finally, we developed a deployable system to deliver translated videos to end users and collect user corrections for iterative improvement.

Retrospective immunophenotypical evaluation of MET, PD-1/PD-L1, and mTOR pathways in primary tumors and pulmonary metastases of renal cell carcinoma: the RIVELATOR study addresses the issue of biomarkers heterogeneity.

In renal cell carcinoma (RCC), tumor heterogeneity generated challenges to biomarker development and therapeutic management, often becoming responsible for primary and acquired drug resistance. This study aimed to assess the inter-tumoral, intra-tumoral, and intra-lesional heterogeneity of known druggable targets in metastatic RCC (mRCC). The RIVELATOR study was a monocenter retrospective analysis of biological samples from 25 cases of primary RCC and their paired pulmonary metastases. The biomarkers analyzed included MET, mTOR, PD-1/PD-L1 pathways and the immune context. High multi-level heterogeneity was demonstrated. MET was the most reliable biomarker, with the lowest intratumor heterogeneity: the positive mutual correlation between MET expression in primary tumors and their metastases had a significantly proportional intensity (P = 0.038). The intratumor heterogeneity grade was significantly higher for the mTOR pathway proteins. Combined immunophenotypical expression patterns and their correlations with the immune context were uncovered [i.e., mTOR expression in the metastases positively correlated with PD-L1 expression in tumor-infiltrating lymphocytes (TILs), P = 0.019; MET expression was related to PD-1 expression on TILs (P = 0.041, ρ = 0.41) and peritumoral lymphocytes (RILs; P = 0.013, ρ = 0.49)], suggesting the possibility of predicting drug response or resistance to tyrosine kinase, mTOR, or immune checkpoint inhibitors. In mRCC, multiple and multi-level assays of potentially predictive biomarkers are needed for their reliable translation into clinical practice. The easy-to-use immunohistochemical method of the present study allowed the identification of different combined expression patterns, providing cues for planning the management of systemic treatment combinations and sequences in an mRCC patient population. The quantitative heterogeneity of the investigated biomarkers suggests that multiple intralesional assays are needed to consider the assessment reliable for clinical considerations.

Test of infant motor performance: Cross-cultural adaptation, validity and reliability in Persian infants

BackgroundThe Test of Infant Motor Performance (TIMP) is an appropriate tool used to measure infant motor performance in many countries and due to the importance of the TIMP in the early diagnosis of infants with developmental delay, the production of a reliable translation of the TIMP in Iran is necessary. AimsThis study was designed to evaluate the validity and reliability of the Persian version of the TIMP and cross-cultural adaptation of the TIMP to Persian. MethodTwenty- two experts were involved in translation, cross-cultural adaptation, validity and reliability. 170 Persian premature infants with post menstrual age (PMA) 34–42 weeks were tested to evaluate inter-rater and intra-rater reliability, test-retest and discriminative validity of the Persian version of the TIMP. ResultsThe content validity and face validity of the items of the Persian version of the TIMP were confirmed by a 10-expert panel and 10 therapists. High intra- and inter-rater reliability (ICC = 0.98, Kappa = 0.93), test-retest reliability (ICC = 0.98) and internal consistency (α = 0.82) were detected for the Persian version of the TIMP. Discriminative validity confirmed that the TIMP could distinguish groups of infants born with a low birth weight (p ˂ 0.001) and infants born small for gestational age (p = 0.002). ConclusionsHigh validity and reliability are reported for the Persian version of the TIMP and this test was demonstrated to be differentiate among groups of infants who are at risk of motor delay.

Development and validation of the Self-Regulated Translation Learning Strategy Scale (SRTLSS)

Self-regulated learning (SRL) has been highlighted as a critical aspect of translation learning. However, past work in translation learning has rarely sought to synthesize translation learning strategies under a self-regulated learning paradigm. Further research would be confined by the paucity of a psychometrically sound instrument to measure learners' self-regulated translation learning strategies. This study aimed to develop a reliable and valid self-regulated translation learning strategy scale (SRTLSS). Based on relevant literature, we hypothesized a hierarchical model and generated items for the SRTLSS. Drawing on the response from 873 Chinese English majors, the validated SRTLSS had 25 items addressing five dimensions: self-control, meaning fulfillment, value and interest, text processing, and retrieving strategies. Results demonstrated the SRTLSS has good measurement validity and reliability with strong invariance across undergraduate and graduate students. The SRTLSS can be used as a complementary tool to diagnose learners' weaknesses and strengths in translation teaching and learning.

Validity and Reliability of Dementia Knowledge Assessment Scale and Dementia Attitude Scale in the Indonesian Language among Nursing Students

BACKGROUND: The population of persons with dementia (PwD) in Indonesia is increasing. As future nurses, it is important to evaluate the knowledge and attitudes of nursing students because they can affect the quality of health services for PwD. However, no instrument is available in the Indonesian language through a good, valid, and reliable translation process. AIM: The aim is to report the translational process and determine the validity and reliability of the Indonesian dementia knowledge assessment scale (DKAS) and dementia attitude scale (DAS) questionnaires in nursing students. METHODS: A quantitative cross-sectional study was conducted involving 486 nursing students from two universities in Yogyakarta, Indonesia. The Brislin method was used to translate DKAS and DAS, then continued with content (involved four experts using items-content validity index [I-CVI]), face, and construct validity tests, as well as reliability tests using Cronbach’s alpha. RESULTS: The I-CVI results were 1.00 for DKAS and 0.98 for DAS, demonstrating a good construct validity index. In addition, the reliability testing had good results, whereas Cronbach’s alpha of the DKAS was 0.713, while DAS was 0.779. CONCLUSION: DKAS and DAS in the Indonesian language are considered valid and reliable to measure nursing student’s knowledge and attitudes toward PwD.

LOVE SONNETS OF GHALIB

"Love Sonnets of Ghalib" is the first and most complete translation of Dewan Ghalib, in which not only the words of Ghalib are translated, but also the description, explanation, phrasing, and translation are specially arranged. This translation is superior to Yusuf Hussain Khan's translation in the sense that it has taken care of all the matters that need clarification, which will help in the interpretation and clarification of Ghalib's thoughts.
 "Urdu Ghazals of Ghalib" was the first organized attempt to properly highlight Ghalib's poetry globally. Which is mentioned by Yusuf Hussain Khan in the foreword And Dr. Abida Begum has mentioned more or less the same thing in the preface, But the fact of the matter is that in this Endeavor, those matters which were necessary for the communication of Ghalib's thought him to non-native environment and civilization were not taken into consideration. However, in "Love Sonnets of Ghalib" a conscious attempt has been made to interpret and evaluate Ghalib's thoughts and ideas. This is a complete translation of Ghalib's popular Diwan. Many examples of Sadeqain's painting are also seen. It was first published in translation in 2002 by Rupa & Co, New Delhi. Its translator is Sarfraz Niazi (1), who is the grandson of Allamah Niaz Fatehpur. Mr Niazi has also informed in the foreword that till now no reliable and authentic translation of Kalam-e-Ghalib is available. That is why the translator chose the way to adopt the literal translation and explain this translation. Here, the poem is translated in two lines only. In determining the value of this translation, some examples will be presented as to how and at what level the translator has represented Ghalib's words and whether all these aspects of Ghalib's interpretation are just due to Sarfaraz Niazi's own efforts. Or someone else's colour is also reflected in these strange images.

Towards Reliable Neural Machine Translation with Consistency-Aware Meta-Learning

Neural machine translation (NMT) has achieved remarkable success in producing high-quality translations. However, current NMT systems suffer from a lack of reliability, as their outputs that are often affected by lexical or syntactic changes in inputs, resulting in large variations in quality. This limitation hinders the practicality and trustworthiness of NMT. A contributing factor to this problem is that NMT models trained with the one-to-one paradigm struggle to handle the source diversity phenomenon, where inputs with the same meaning can be expressed differently. In this work, we treat this problem as a bilevel optimization problem and present a consistency-aware meta-learning (CAML) framework derived from the model-agnostic meta-learning (MAML) algorithm to address it. Specifically, the NMT model with CAML (named CoNMT) first learns a consistent meta representation of semantically equivalent sentences in the outer loop. Subsequently, a mapping from the meta representation to the output sentence is learned in the inner loop, allowing the NMT model to translate semantically equivalent sentences to the same target sentence. We conduct experiments on the NIST Chinese to English task, three WMT translation tasks, and the TED M2O task. The results demonstrate that CoNMT effectively improves overall translation quality and reliably handles diverse inputs.

Corpus-based Intangible Cultural Heritage Translation Path and Strategy Analysis of Chinese Culture "Going Out"

Intangible cultural heritage translation is regarded as a new subject in the new era of economic construction development, can not only the external publicity of the rich national culture, but also improve our country's cultural propaganda comprehensive level. By studying the corpus-based intangible cultural heritage translation path, the quality and efficiency of cultural translation can be comprehensively improved, and the reliable translation principles and effective strategies of large-scale data can be defined, which can provide effective basis for international cultural publicity in the new era. Therefore, on the basis of understanding the research status of corpus and intangible cultural heritage translation, this paper mainly discusses the corpus-based machine translation system, and then analyzes effective strategies for Chinese culture to "go out" from the perspective of practical development.