Automatic Translation Research Articles

The Development of AI-Powered Automatic Video Sign Language Translation Systems

People who are deaf or hard of hearing often face challenges in fully understanding written subtitles in videos due to the differences between sign language and spoken language, each having unique grammar and structures. While many social media platforms provide automatic subtitles, they are often insufficient for accessibility. Sign language interpretation should be included to make video content fully accessible. Traditionally, adding sign language to videos involves a time-consuming process of recording and embedding separate translation videos, which must be redone with any changes to the original video. The plugin simplifies this process by dynamically translating updates directly from subtitle files, significantly reducing the time, effort, and cost involved. It allows seamless sign language support for any video without modifying the original content. The plugin integrates with video players, providing a user-controlled, customizable sign language window that can be activated, moved, resized, or turned off as desired.

Life values in Russia and China: Semantic analysis and comparison

The paper seeks to explore the core life values of people living in Russia and China using an online survey in the respondents’ native languages with samples of 887 and 559 participants from each country respectively, who were asked to think of three words associated ‘with something most important in life’. The word-association arrays were translated into English by an automatic translator; then we compared the frequencies of individual words or those of their meanings obtained using the Roget’s Thesaurus. The words ‘family’ and ‘love’ were the most frequent in both samples, but the frequencies of these and most of the other words from the semantic core of each sample differed significantly. Analyses at the level of semantic vectors showed that more than half of the meanings that Russian respondents described as important and more than a third of the meanings important to the Chinese respondents had significant differences in frequencies between the samples. The paper presents 62 semantic groups of the Roget’s Thesaurus, the frequencies of which are reliably different; these are the meanings in which the Chinese and Russian samples differ with regard to life values. There are also 26 semantic groups, the frequencies of which are statistically the same and comparatively high in both samples: these are the common meanings. The results may be used in the design of automatic and AI systems to reduce the intercultural barriers.

SWL-LSE: A Dataset of Health-Related Signs in Spanish Sign Language with an ISLR Baseline Method

Progress in automatic sign language recognition and translation has been hindered by the scarcity of datasets available for the training of machine learning algorithms, a challenge that is even more acute for languages with smaller signing communities, such as Spanish. In this paper, we introduce a dataset of 300 isolated signs in Spanish Sign Language, collected online via a web application with contributions from 124 participants, resulting in a total of 8000 instances. This dataset, which is openly available, includes keypoints extracted using MediaPipe Holistic. The goal of this paper is to describe the construction and characteristics of the dataset and to provide a baseline classification method using a spatial–temporal graph convolutional network (ST-GCN) model, encouraging the scientific community to improve upon it. The experimental section offers a comparative analysis of the method’s performance on the new dataset, as well as on two other well-known datasets. The dataset, code, and web app used for data collection are freely available, and the web app can also be used to test classifier performance on-line in real-time.

« Vous pouvez entendre les mâchoires de l’arme de l’ange ». Le traducteur humain face à la traduction automatique dans le contexte littéraire : Les Livres de Jakób d’Olga Tokarczuk

The dynamic development of automatic translators raises questions about the future of human translators. The aim of this article is to evaluate the French versions of an excerpt from The Books of Jacob by Olga Tokarczuk, made by Systran, Google Translate and DeepL, and to compare them with the reference translation made by a human translator. The translations were carried out in 2019 as part of the ATLAS association project ‘L’Observatoire de la traduction automatique’. The analyses show that the inability to interpret the original in a complex way leads to weaknesses in machine translators. This proves that the position of the human translator remains unthreatened in the field of translating high-quality literary texts.

Research on Automatic Proofreading Algorithm for English Translation Based on Neural Networks

In this proposed study, we explore the development and implementation of an innovative proofreading algorithm aimed at enhancing the accuracy of English translation. This algorithm leverages the capabilities of Convolutional Neural Networks (CNN) integrated with a fuzzy logic approach, offering a novel perspective in the realm of linguistic accuracy and consistency in translations. The core objective of this research is to address the prevalent challenges in automatic translation, such as context misinterpretation and semantic errors, by employing a fuzzy-based CNN model. This model is meticulously trained and tested using a diverse dataset of English translations, enabling it to learn and adapt to various linguistic nuances. Our results demonstrate a significant improvement in the proofreading accuracy, outperforming existing methods in terms of efficiency and reliability. The research highlights the potential of combining neural networks with fuzzy logic to create more sophisticated and context-aware translation tools. While our findings mark a considerable advancement in automatic translation proofreading, we also acknowledge the scope for further enhancements. Future work could involve refining the algorithm, expanding its applicability to other languages, and integrating it into real-world translation software. This research contributes to the evolving landscape of automated translation, presenting a promising solution for achieving higher translation fidelity.

Automatic sign language translation: a review of neural network methods for recognition and synthesis of spoken and signed language

Automatic sign language translation: a review of neural network methods for recognition and synthesis of spoken and signed language

An automatic merging method for large kinetic mechanisms: Application to a surrogate fuel combustion model

Comparing and merging complex kinetic models developed by different teams is a delicate operation, prone to errors, not least because of the multiplicity of usages for naming and describing chemical species. However, the modelling of combustion systems calls for increasingly complex kinetics, especially for mixtures, which can often be built from different mechanisms developed for single components. The TIRAMISU automatic fusion code, written in Python, has been developed for an efficient automatic merging of large mechanisms. This new code proceeds in several stages. A preliminary reading and cleaning of each mechanism is followed by an automatic translation step based on the comparison of the Gibbs energy of the isomers to be matched, as well as the reactions involving them. Two analyses of possible matches between isomers from the two mechanisms are proposed, based on thermodynamic properties. When the matches are not sufficient for a confident translation, the system uses the similarities of the reactions involving the species. Finally, a model merging step is proposed to the user, which can be total or limited to the sub-mechanism of a species taken from one mechanism to be added to another. These two objectives are illustrated by the construction of a TRF model (toluene, n-heptane, isooctane blend surrogate of gasoline) by total fusion of two models from the literature. The possibility of choosing thermochemical and kinetics data from either of the parent models is illustrated, as is the impact on simulations of auto-ignition times. This TRF model is then extended with TIRAMISU to a TRF-iB by including the sub-mechanism of a bio-fuel, isobutanol, from other models in the literature. The impact of the user’s choice of the size of the sub-mechanism on simulation fidelity and the size of the resulting model is discussed, illustrating the power and versatility of this new software.

Neural Machine Translation Between English And Bangla: A Philosophical Survey Observing Architectures And Performances

Neural Machine Translation (NMT) is a state of the art of machine translations that uses neural network models to translate text from one language to another. Unlike traditional machine translation methods, which rely on statistical models and rule-based systems, NMT leverages the power of deep learning, specifically using techniques such as recurrent neural networks (RNNs), long short-term memory networks (LSTMs), or more recently, transformers. Neural Machine Translation represents a significant dive forward in the field of automatic translation, leveraging the latest advancements in deep learning to deliver more accurate and natural translations. The progress in NMT has been remarkable, driven by advancements in model architectures, multilingual training, pretraining techniques, increased computational power, and the availability of vast multilingual datasets and improved evaluation metrics. These developments have collectively enhanced the quality, efficiency, and applicability of NMT systems across various languages and domains. In this article we will explore the technologies and performances of NMT between English and Bangla Language pair from its launching periods to present.

DiBiMT: A Gold Evaluation Benchmark for Studying Lexical Ambiguity in Machine Translation

Abstract Despite the remarkable progress made in the field of Machine Translation (MT), current systems still struggle when translating ambiguous words, especially when these express infrequent meanings. In order to investigate and analyze the impact of lexical ambiguity on automatic translations, several tasks and evaluation benchmarks have been proposed over the course of the last few years. However, works in this research direction suffer from critical shortcomings. Indeed, existing evaluation datasets are not entirely manually curated, which significantly compromises their reliability. Furthermore, current literature fails to provide detailed insights into the nature of the errors produced by models translating ambiguous words, lacking a thorough manual analysis across languages. With a view to overcoming these limitations, we propose Disambiguation Biases in MT (DiBiMT), an entirely manually-curated evaluation benchmark for investigating disambiguation biases in eight language combinations and assessing the ability of both commercial and non-commercial systems to handle ambiguous words. We also examine and detail the errors produced by models in this scenario by carrying out a manual error analysis in all language pairs. Additionally, we perform an extensive array of experiments aimed at studying the behavior of models when dealing with ambiguous words. Finally, we show the ineffectiveness of standard MT evaluation settings for assessing the disambiguation capabilities of systems and highlight the need for additional efforts in this research direction and ad-hoc testbeds such as DiBiMT.Our benchmark is available at: https://nlp.uniroma1.it/dibimt/.

Construction of English corpus oral instant translation model based on information security and deep learning

In order to improve the quality of translation, avoid translation ambiguity and accurately present the content of the source language, supported by the concept of deep learning and guaranteed by information security, an instant oral translation model is constructed for English corpus. The aim of this study is to enhance the efficiency and accuracy of oral translation systems through the application of deep learning algorithms. Specifically, we employ a sample training mechanism tailored to the unique characteristics of oral translation, allowing for separate training of system interaction and translation data. Furthermore, by redesigning the interaction hardware, this research comprehensively redefines the hardware structure of the translation system, marking a significant step towards improving the usability and performance of such systems. After obtaining and processing effective security sensitive information, language resources are managed by using database management system, which fundamentally improves the level of network information security. The performance of the existing oral automatic translation system (Test Group 1) and the system designed in this paper (Test Group 2) is tested by experiments, and the results are as follows: (1) The translation system designed here has better interactive performance, and it is better than Test Group 1. (2) The adaptive index value of Test Group 1 is 1, and that of Test Group 2 is 0.5, which proves that the adaptive ability of system algorithm of Test Group 2 is better than that of Test Group 1. (3) When comparing the translation speed, the translation time of Test Group 2 is only 70.7 s, while that of Test Group 1 is 130.6 s, so the proposed translation system is obviously superior to that of Test Group 1.

Adapting Machine Translation Engines to the Needs of Cultural Heritage Metadata

The Europeana digital library features cultural heritage collections from over 3,000 European institutions described in 37 languages. However, most textual metadata describe the records in a single language, the data providers’ language. Improving Europeana’s multilingual accessibility presents challenges due to the unique characteristics of cultural heritage metadata, often expressed in short phrases and using in-domain terminology. This work presents the EuropeanaTranslate project’s approach and results, aimed at translating Europeana metadata records from 23 EU languages into English. Machine Translation engines were trained on a cleaned selection of bilingual and synthetic data from Europeana, including multilingual vocabularies and relevant cultural heritage repositories. Automatic translations were evaluated through standard metrics and human assessments by linguists and domain cultural heritage experts. The results showed significant improvements when compared to the generic engines used before the in-domain training as well as the eTranslation service for most languages. The EuropeanaTranslate engines have translated over 29 million metadata records on Europeana.eu. Additionally, the MT engines and training datasets are publicly available via the European Language Grid Catalogue and the ELRC-SHARE repository.

Lobish: Symbolic Language for Interpreting Electroencephalogram Signals in Language Detection Using Channel-Based Transformation and Pattern.

Electroencephalogram (EEG) signals contain information about the brain's state as they reflect the brain's functioning. However, the manual interpretation of EEG signals is tedious and time-consuming. Therefore, automatic EEG translation models need to be proposed using machine learning methods. In this study, we proposed an innovative method to achieve high classification performance with explainable results. We introduce channel-based transformation, a channel pattern (ChannelPat), the t algorithm, and Lobish (a symbolic language). By using channel-based transformation, EEG signals were encoded using the index of the channels. The proposed ChannelPat feature extractor encoded the transition between two channels and served as a histogram-based feature extractor. An iterative neighborhood component analysis (INCA) feature selector was employed to select the most informative features, and the selected features were fed into a new ensemble k-nearest neighbor (tkNN) classifier. To evaluate the classification capability of the proposed channel-based EEG language detection model, a new EEG language dataset comprising Arabic and Turkish was collected. Additionally, Lobish was introduced to obtain explainable outcomes from the proposed EEG language detection model. The proposed channel-based feature engineering model was applied to the collected EEG language dataset, achieving a classification accuracy of 98.59%. Lobish extracted meaningful information from the cortex of the brain for language detection.

SPECIALIZED NEURAL MACHINE TRANSLATION FOR AVIATION ENGLISH: INSIGHTS FROM PORTUGUESE-TO-ENGLISH AVIATION SAFETY REPORTS

Aviation safety reports are essential sources for the identification and analysis of risks in civil aviation. These reports are written in plain language, which requires the application of Natural Language Processing techniques for automatic and intelligent treatment. In the case of Brazil, the vast majority of reports are written in Portuguese. Therefore, for comparison with international database of reports that are written in English, a first step is the translation of Brazilian reports. In this work, a proposal for a machine translation model is presented based on the fine-tuning of pre-trained models. To this end, an aviation-specific language corpus is developed with the objective of generating example data for model training. Finally, a pre-trained model is fine-tuned with the corpus created in order to implement an automatic translation model that achieves good results in the task considering the specifics of the aviation language. As a result, a first model is implemented, presenting coherent results of translation between Portuguese/English in the specific domain of aviation.

Revitalize Hearing Skills with Advanced Speech Processing Technology

Revitalizing listening skills with advanced sound processing technology is an interesting concept, especially in today's digital era where technology is advancing rapidly. Listening skill is the ability to understand, analyze, and respond well to information conveyed through sound, and this is a key skill in various contexts, such as interpersonal communication, business, education, and others. Here are some ways revitalizing hearing skills with advanced sound processing technologies can be accomplished, Introduction to Sound Processing Technologies, It is important to understand what advanced sound processing technologies are. This includes using speech processing algorithms, speech recognition techniques, and even artificial intelligence (AI) to optimize speech understanding and interpretation. Hearing Skills Training Speech processing technology can be used in hearing skills training. For example, an application that can record, analyze and provide feedback on a person's hearing ability. Such applications can be used in the context of business education or training. Automatic Transcription and Translation, Voice processing technology can be used to convert conversations in foreign languages ??into text or translate voice content in real time. This can help someone understand a foreign language or content displayed in a language they are not familiar with. Improved Business Communications, In business, voice processing technology can be used to record, analyze and provide feedback on presentations, meetings or telephone conversations. This can help improve business communication skills. Providing Better Customer Service Many companies use voice processing technology to improve their customer service. Advanced voice interactive systems (IVR) can be used to route customer calls more efficiently and provide necessary information. Sentiment Understanding Advanced voice processing technology can be used to analyze sentiment in conversations or customer reviews. This can help companies understand the views and feelings of their customers. Educational Applications Speech processing technology can also be used in education to create more interactive teaching materials.

Open Brain AI and language assessment.

Neurolinguistic assessments play a vital role in neurological examinations, revealing a wide range of language and communication impairments associated with developmental disorders and acquired neurological conditions. Yet, a thorough neurolinguistic assessment is time-consuming and laborious and takes valuable resources from other tasks. To empower clinicians, healthcare providers, and researchers, we have developed Open Brain AI (OBAI). The aim of this computational platform is twofold. First, it aims to provide advanced AI tools to facilitate spoken and written language analysis, automate the analysis process, and reduce the workload associated with time-consuming tasks. The platform currently incorporates multilingual tools for English, Danish, Dutch, Finnish, French, German, Greek, Italian, Norwegian, Polish, Portuguese, Romanian, Russian, Spanish, and Swedish. The tools involve models for (i) audio transcription, (ii) automatic translation, (iii) grammar error correction, (iv) transcription to the International Phonetic Alphabet, (v) readability scoring, (vi) phonology, morphology, syntax, semantic measures (e.g., counts and proportions), and lexical measures. Second, it aims to support clinicians in conducting their research and automating everyday tasks with "OBAI Companion," an AI language assistant that facilitates language processing, such as structuring, summarizing, and editing texts. OBAI also provides tools for automating spelling and phonology scoring. This paper reviews OBAI's underlying architectures and applications and shows how OBAI can help professionals focus on higher-value activities, such as therapeutic interventions.

Machine Translation for Open Scholarly Communication: Examining the Relationship between Translation Quality and Reading Effort

This study assesses the usability of machine-translated texts in scholarly communication, using self-paced reading experiments with texts from three scientific disciplines, translated from French into English and vice versa. Thirty-two participants, proficient in the target language, participated. This study uses three machine translation engines (DeepL, ModernMT, OpenNMT), which vary in translation quality. The experiments aim to determine the relationship between translation quality and readers’ reception effort, measured by reading times. The results show that for two disciplines, manual and automatic translation quality measures are significant predictors of reading time. For the most technical discipline, this study could not build models that outperformed the baseline models, which only included participant and text ID as random factors. This study acknowledges the need to include reader-specific features, such as prior knowledge, in future research.

Translation as a linguistic act in the context of artificial intelligence: the impact of technological changes on traditional approaches

The purpose of this article is to study translation as a human speech act in the context of artificial intelligence. Using the method of analysing the related literature, the article focuses on the impact of technological changes on traditional approaches and explores the links between these concepts and their emergence in linguistics and automatic language processing methods. The results show that the main methods include stochastic, rule-based, and methods based on finite automata or expressions. Studies have shown that stochastic methods are used for text labelling and resolving ambiguities in the definition of word categories, while contextual rules are used as auxiliary methods. It is also necessary to consider the various factors affecting automatic language processing and combine statistical and linguistic methods to achieve better translation results. Conclusions - In order to improve the performance and efficiency of translation systems, it is important to use a comprehensive approach that combines various techniques and machine learning methods. The research confirms the importance of automated language processing in the fields of AI and linguistics, where statistical methods play a significant role in achieving better results. Keywords: technological changes, linguistics, innovations, language technologies, automatic translation

Whisper-SV: Adapting Whisper for low-data-resource speaker verification

Trained on 680,000 h of massive speech data, Whisper is a multitasking, multilingual speech foundation model demonstrating superior performance in automatic speech recognition, translation, and language identification. However, its applicability in speaker verification (SV) tasks remains unexplored, particularly in low-data-resource scenarios where labeled speaker data in specific domains are limited. To fill this gap, we propose a lightweight adaptor framework to boost SV with Whisper, namely Whisper-SV. Given that Whisper is not specifically optimized for SV tasks, we introduce a representation selection module to quantify the speaker-specific characteristics contained in each layer of Whisper and select the top-k layers with prominent discriminative speaker features. To aggregate pivotal speaker-related features while diminishing non-speaker redundancies across the selected top-k distinct layers of Whisper, we design a multi-layer aggregation module in Whisper-SV to integrate multi-layer representations into a singular, compacted representation for SV. In the multi-layer aggregation module, we employ convolutional layers with shortcut connections among different layers to refine speaker characteristics derived from multi-layer representations from Whisper. In addition, an attention aggregation layer is used to reduce non-speaker interference and amplify speaker-specific cues for SV tasks. Finally, a simple classification module is used for speaker classification. Experiments on VoxCeleb1, FFSVC, and IMSV datasets demonstrate that Whisper-SV achieves EER/minDCF of 2.22%/0.307, 6.14%/0.488, and 7.50%/0.582, respectively, showing superior performance in low-data-resource SV scenarios.

Automatic translation from English to Amazigh using transformer learning

Due to the lack of parallel data, to our knowledge, no study has been conducted on the Amazigh-English language pair, despite the numerous machine translation studies completed between major European language pairs. We decided to utilize the neural machine translation (NMT) method on a parallel corpus of 137,322 sentences. The attention-based encoder-decoder architecture is used to construct statistical machine translation (SMT) models based on Moses, as well as NMT models using long short-term memory (LSTM), gated recurrent units (GRU), and transformers. Various outcomes were obtained for each strategy after several simulations: 80.7% accuracy was achieved using the statistical approach, 85.2% with the GRU model, 87.9% with the LSTM model, and 91.37% with the transformer.

SIMILARITIES OF LEXICAL-SEMANTIC RELATIONS IN UZBEK AND ENGLISH LANGUAGES

This article deals with corpus linguistics, ideas about the corpus and its parallel corpus link, its structure, corpus types, tokens, lemmas, stemming. Today, the theoretical and practical significance of the corps is in the study of the existing possibilities of language in Uzbek linguistics, the identification of problematic aspects of linguistics, the creation of electronic dictionaries, increasing the effectiveness of modern information technology in language learning, automatic translation, search and computer analysis. In solving problems, there is a need to build a corpus of language in specific areas.