Automatic Translation Research Articles

Using Symmetries to Lift Satisfiability Checking

We analyze how symmetries can be used to compress structures (also known as interpretations) onto a smaller domain without loss of information. This analysis suggests the possibility to solve satisfiability problems in the compressed domain for better performance. Thus, we propose a 2-step novel method: (i) the sentence to be satisfied is automatically translated into an equisatisfiable sentence over a ``lifted'' vocabulary that allows domain compression; (ii) satisfiability of the lifted sentence is checked by growing the (initially unknown) compressed domain until a satisfying structure is found. The key issue is to ensure that this satisfying structure can always be expanded into an uncompressed structure that satisfies the original sentence to be satisfied. We present an adequate translation for sentences in typed first-order logic extended with aggregates. Our experimental evaluation shows large speedups for generative configuration problems. The method also has applications in the verification of software operating on complex data structures. Our results justify further research in automatic translation of sentences for symmetry reduction.

Assume-Guarantee Reinforcement Learning

We present a modular approach to reinforcement learning (RL) in environments consisting of simpler components evolving in parallel. A monolithic view of such modular environments may be prohibitively large to learn, or may require unrealizable communication between the components in the form of a centralized controller. Our proposed approach is based on the assume-guarantee paradigm where the optimal control for the individual components is synthesized in isolation by making assumptions about the behaviors of neighboring components, and providing guarantees about their own behavior. We express these assume-guarantee contracts as regular languages and provide automatic translations to scalar rewards to be used in RL. By combining local probabilities of satisfaction for each component, we provide a lower bound on the probability of satisfaction of the complete system. By solving a Markov game for each component, RL can produce a controller for each component that maximizes this lower bound. The controller utilizes the information it receives through communication, observations, and any knowledge of a coarse model of other agents. We experimentally demonstrate the efficiency of the proposed approach on a variety of case studies.

Enhancing Industrial Digitalisation through an Adaptable Component for Bridging Semantic Interoperability Gaps

Digital transformation has a great effect on industries but also in the perception of their products with high consequences on society as a whole, as driver for innovation. However, there are still many barriers to the adoption of digital technologies, especially in small and medium-sized enterprises (SMEs). In the present study we propose a new software component that facilitates the adoption of industrial reference architectures (RAs) based on the semantic interoperability concept. Among these RAs, FIWARE has become a key enabler for flexible and resource effective digitalisation of shop floor processes. The proposed software component aims to enable full interoperability between these more widely used technologies and industrial standards and protocols such as OPC UA. As a result, a flexible, open, and adaptable implementation assists the automatic translation of messages from different standards enabling semantic interoperability of shop floor processes. The component aims to enable industrial SMEs to bring current barriers to digitalisation towards a more human-centred, resilient, and sustainable industry and society.

SEMANTIC LABELING OF 3D BUILDINGS BY USING GRAPH NEURAL NETWORK (GNN)

Abstract. Understanding building semantics is crucial for comprehending their structure, components, and functions. This study investigates the utilization of Graph Neural Networks (GNN) to semantically label 3D building models, aiming to optimize these labels' output within Geographic Information Systems (GIS). The methodology leverages diverse datasets comprising 3D building models, utilizing BuildingGNN algorithms to iteratively refine the labeling process. The resultant labeled components, after thorough validation, are translated into the CityGML format. The generated CityGML dataset holds promise for a wide array of 3D city model applications, heightening the utility of labeled components within GIS analyses. While exploring GNN's capabilities may reduce manual effort and time and provide standardized data representation, this study also addresses challenges in ensuring output reliability. While automation remains a goal, future research endeavors may focus on refining automatic labeling techniques and semantic translation processes to further improve accuracy and applicability.

Impact of Language-Specific Training on Image Caption Synthesis: A Case Study on Low-Resource Assamese Language

Automatic Image Captioning (AIC) refers to the process of synthesizing semantically and syntactically correct descriptions for images. Existing research on AIC has predominantly focused on the English language. Comparatively, lower numbers of works have focused on developing captioning systems for low-resource Indian languages like Assamese. This paper investigates AIC for the Assamese language using two distinct approaches. The first approach involves utilizing state-of-the-art AIC model pretrained on an English image-caption dataset to generate English captions for input images. Next, these English captions are translated to the Assamese language using a publicly available automatic translator. The second approach involves exclusively training the AIC model using an Assamese image-caption dataset to predict captions directly in Assamese. The experiments are performed on two types of state-of-art models, one which uses LSTM as a decoder and the other one uses a transformer. Through extensive experimentation, the performance of these approaches is evaluated both quantitatively and qualitatively. The quantitative results are obtained using automatic metrics such as BLEU-n and CIDEr. For qualitative analysis, human evaluation is performed. The comparative performances between the two approaches reveal that models trained exclusively on Assamese image-caption datasets achieve superior results both in terms of quantitative measures and qualitative assessment when compared to models pretrained on English and subsequently translated into Assamese.

The Role of Avatars in Language Learning in the Metaverse

In the field of education, new technologies have given rise to innovative teaching methodologies to cater to digital natives. Among them, the metaverse and its emphasis on multimodality allow students to embody avatars to explore replicas of the physical world. This phenomenon democratizes access to cultural and social experiences through teleportation, challenging the limitations of the physical world. For all this to be possible, the avatar, central to this dynamic, acts as an intermediary between the metaverse and the real world, fostering three-dimensional cooperation and global interaction through automatic translation and accessibility features. Specifically, in the realm of language teaching, the avatar represents an alter ego that promotes the ideal and polysemic self, reducing social pressure and providing security to the student. On the other hand, in the learning process, the avatar promotes student autonomy, allowing for a constructivist and communicative approach. The recreation of diverse linguistic and extralinguistic contexts in the metaverse enables students to develop language and cultural competencies in an immersive manner. Furthermore, the sharing and sensory aspects of scenarios and teaching materials contribute to richer feedback, strengthening the sense of learning and student security in this three-dimensional, barrier-free environment.

MedLingua: A conceptual framework for a multilingual medical conversational agent

This study introduces a hybrid model for an advanced medical chatbot addressing crucial healthcare communication challenges. Leveraging a hybrid ML model, the chatbot aims to provide accurate and prompt responses to users' health-related queries. The proposed model will overcome limitations observed in previous medical chatbots by integrating a dual-stemming approach, P-Stemmer and NLTK-Stemmer, accommodating both semitic and non-semitic languages. The system prioritizes the analysis of cognates, identification of symptoms, doctor recommendations, and prescription generation. It integrates an automatic translation module to facilitate a smooth multilingual diagnostic experience. Following the Scrum methodology for agile development, the framework ensures adaptability to evolving research needs and stays current with recent medical discoveries. This groundbreaking idea aims to improve the effectiveness and availability of healthcare services by introducing an intelligent, multilingual chatbot. This technology enables patients to communicate with doctors from diverse linguistic backgrounds through an automated language translation model, eliminating language barriers and extending healthcare access to rural regions worldwide.•A simple but efficient hybrid conceptual model for advancement in smart medical assistance.•This conceptual model can be applied to implement a medical chatbot that can understand multiple languages.•This method can be utilized to address medical chatbot limitations and enhance accuracy in response generation.

A STUDY OF GRAMMATICAL ERRORS WHEN USING TRANSLATION SOFTWARE FOR STUDENTS IN ECONOMIC CONTRACTS

Along with the strong development of the ongoing fourth industrial revolution, automatic translation software also plays a very important role in helping people get closer together without being hindered by language. Google Translate software is one of the early software and is gradually improving itself. The birth of this software has helped people who do not know each other's languages to communicate and read documents to get closer. However, the correct translation is still a problem due to many reasons. In order to clarify grammatical errors when using google translate software to translate from English to Vietnamese, the author will conduct a study on the situation in economic contracts through 100 English contracts collected from foreign companies. cooperation between vnguowif Vietnam and foreigners. The purpose of this study is to point out common mistakes and suggest ways to prevent students from using the software. Besides, it also helps software developers have a scientific basis to perfect it in the future.

A comparative analysis of lexical-based automatic evaluation metrics for different Indic language pairs

<p>With the rise of machine translation systems, it has become essential to evaluate the quality of translations produced by these systems. However, the existing evaluation metrics designed for English and other European languages may not always be suitable or apply to other Indic languages due to their complex morphology and syntax. Machine translation evaluation (MTE) is a process of assessing the quality and accuracy of the machine-translated text. MTE involves comparing the machine-translated output with the reference translation to calculate the level of similarity and correctness. Therefore, this study evaluates different metrics, namely, BLEU, METEOR, and TER to identify the most suitable evaluation metric for Indic languages. The study uses datasets for Indic languages and evaluates the metrics on various translation systems. The study contributes to the field of MT by providing insights into suitable evaluation metrics for Indic languages. This research paper aims to study and compare several lexical automatic machine translation evaluation metrics for Indic languages. For this research analysis, we have selected five language pairs of parallel corpora from the low-resource domain, such as English–Hindi, English-Punjabi, English-Gujarati, English-Marathi, and English-Bengali. All these languages belong to the Indo-Aryan language family and are resource-poor. A comparison of the state of art MT is presented and shows which translator works better on these language pairs. For this research work, the natural language toolkit tokenizers are used to assess the analysis of the experimental results. These results have been performed by taking two different datasets for all these language pairs using fully automatic MT evaluation metrics. The research study explores the effectiveness of these metrics in assessing the quality of machine translations between various Indic languages. Additionally, this dataset and analysis will make it easier to do future research in Indian MT evaluation.</p>

Spoken‐to‐written text conversion for enhancement of Korean–English readability and machine translation

AbstractThe Korean language has written (formal) and spoken (phonetic) forms that differ in their application, which can lead to confusion, especially when dealing with numbers and embedded Western words and phrases. This fact makes it difficult to automate Korean speech recognition models due to the need for a complete transcription training dataset. Because such datasets are frequently constructed using broadcast audio and their accompanying transcriptions, they do not follow a discrete rule‐based matching pattern. Furthermore, these mismatches are exacerbated over time due to changing tacit policies. To mitigate this problem, we introduce a data‐driven Korean spoken‐to‐written transcription conversion technique that enhances the automatic conversion of numbers and Western phrases to improve automatic translation model performance.

Re-thinking Machine Translation Post-Editing Guidelines

Machine Translation Post-Editing (MTPE) is a challenging task. It frequently creates tension between what the industry expects in terms of quality and what translators are willing to deliver as an end product. Conventional approaches to MTPE take as a point of departure the distinction between light and full MPTE, but the division gets blurred when implemented in an actual MTPE project where translators find difficulties in differentiating between essential and preferential changes. At the time MTPE guidelines were designed, the role of the human translator in the MT process was perceived as ancillary, a view inherited from the first days of MT research aiming at the so-called Fully Automatic High Quality Machine Translation (FAHQMT). My proposal challenges the traditional division of MTPE levels and presents a new way of looking at MTPE guidelines. In view of the latest developments in neural machine translation and the higher quality level of its output, it is my contention that the traditional division of MTPE levels is no longer valid. In this contribution I advance a proposal for redefining MTPE guidelines in the framework of an ecosystem specifically designed for this purpose.

GPT-Driven Source-to-Source Transformation for Generating Compilable Parallel CUDA Code for Nussinov’s Algorithm

Designing automatic optimizing compilers is an advanced engineering process requiring a great deal of expertise, programming, testing, and experimentation. Maintaining the approach and adapting it to evolving libraries and environments is a time-consuming effort. In recent years, OpenAI has presented the GPT model, which is designed for many fields like computer science, image processing, linguistics, and medicine. It also supports automatic programming and translation between programming languages, as well as human languages. This article will verify the usability of the commonly known LLM model, GPT, for the non-trivial NPDP Nussinov’s parallel algorithm code within the OpenMP standard to create a parallel equivalent of CUDA for NVIDIA graphics cards. The goal of this approach is to avoid creating any post-processing scripts and writing any lines of target code. To validate the output code, we compare the resulting arrays with the ones calculated by the optimized code for the CPU generated employing the polyhedral compilers. Finally, the code will be checked for scalability and performance. We will concentrate on assessing the capabilities of GPT, highlighting common challenges that can be refined during future learning processes. This will enhance code generation for various platforms by leveraging the outcomes from polyhedral optimizers.

Regularizing cross-attention learning for end-to-end speech translation with ASR and MT attention matrices

The cross-attention mechanism enables Transformer to capture correspondences between the input and output. However, in the domain of end-to-end (E2E) speech-to-text translation (ST), the learned cross-attention weights often struggle to accurately correspond with actual alignments, given the need to align speech and text across different modalities and languages. In this paper, we present a simple yet effective method called regularized cross-attention learning, for end-to-end speech translation in a multitask learning (MTL) framework. RCAL leverages the knowledge from auxiliary automatic speech recognition (ASR) and machine translation (MT) tasks to generate a teacher cross-attention matrix, serving as prior alignment knowledge to enhance cross-attention learning within the ST task. An additional loss function is introduced as part of the MTL framework to facilitate this process. We conducted experiments on the MuST-C benchmark dataset to evaluate the effectiveness of RCAL. The results demonstrate that the proposed approach yields significant improvements over the baseline, with an average enhancement of +0.8 BLEU across four translation directions in two experimental settings, outperforming state-of-the-art E2E and cascaded speech translation models. Further analysis and visualization reveal that the model with RCAL effectively learns high-quality alignment information from auxiliary ASR and ST tasks, thereby improving the ST alignment quality. Moreover, the experiments with different sizes of MT and ST data provide strong evidence supporting our model’s robustness in various scenarios.

Automatic translation of human route descriptions into schematic maps for indoor navigation

ABSTRACT People create route descriptions based on their mental maps to provide route guidance, which represents their knowledge of the environment. Recent studies have attempted to model navigation knowledge from human route descriptions to facilitate route communication. However, they mainly focus on outdoor environments and do not address the representation of human descriptions of indoors in the form of schematic maps through the automatic extraction of spatial knowledge. Schematic maps have been commonly applied for public transportation by utilizing abstract representations to reduce cognitive load. Compared to route descriptions, schematic maps can provide easy-to-understand navigation guidance. In this paper, we present a novel NLP-based pipeline to automatically generate schematic maps from human route descriptions for indoor navigation. The experimental data consists of a set of crowdsourced route descriptions that follow a common template for a test building of the Soleway indoor navigation web service. The route descriptions and the generated schematic maps were presented to human participants in an online survey, and it was found that 92% of the generated schematic maps matched well with the corresponding human route descriptions. Thus, the proposed method is an effective and reliable approach for modeling route descriptions through schematic maps in indoor route communication.

A TinyML solution for an IoT-based communication device for hearing impaired

Research on automatic translation of sign language to verbal languages has been progressively explored in recent years to assist speech and hearing-impaired people in communicating with non-signers. In this paper, a tiny machine learning (TinyML) solution is proposed for sign language recognition using a low-cost, wearable, internet-of things (IoT) device. A lightweight deep neural network is deployed on the edge device to interpret isolated signs from the Indian sign language using the time-series data collected from the motion sensors of the device. The scarcity of labeled training data is addressed by employing the deep transfer learning approach. Here, the knowledge gained from the data collected using the motion sensors of a different device is used to initialize the model parameters. The performance of the model is assessed in terms of classification accuracy and prediction time for different sampling rates and transferring schemes. The model achieves an average accuracy of 87.18% when all the parameters are retrained with just 4 observations of each sign recorded from the motion sensors of the proposed IoT device. The recognized sign is transmitted to a cloud platform in real-time. A mobile application, SignTalk, is also developed, which wirelessly receives the predicted signs from the cloud and displays it as text. Additionally, text-to-speech conversion is also provided on SignTalk to vocalize the predicted sign for better communication.

Enhancing Transparency in Defining Studied Drugs: The Open-Source Living DiAna Dictionary for Standardizing Drug Names in the FAERS.

In refining drug safety signals, defining the object of study is crucial. While research has explored the effect of different event definitions, drug definition is often overlooked. The US FDA Adverse Event Reporting System (FAERS) records drug names as free text, necessitating mapping to active ingredients. Although pre-mapped databases exist, the subjectivity and lack of transparency of the mapping process lead to a loss of control over the object of study. We implemented the DiAna dictionary, systematically mapping individual free-text instances to their corresponding active ingredients and linking them to the World Health Organization Anatomical Therapeutic Chemical (WHO-ATC) classification. We retrieved all drug names reported to the FAERS (2004-December 2022). Using existing vocabularies and string editing, we automatically mapped free text to ingredients. We manually revised the mapping and linked it to the ATC classification. We retrieved 18,151,842 reports, with 74,143,411 drug entries. We manually checked the first 14,832 terms, up to terms occurring over 200 times (96.88% of total drug entries), to 6282 unique active ingredients. Automatic unchecked translations extend the standardization to 346,854 terms (98.94%). The DiAna dictionary showed a higher sensitivity compared with RxNorm alone, particularly for specific drugs (e.g., rimegepant, adapalene, drospirenone, umeclidinium). The most prominent drug classes in the FAERS were immunomodulating (37.40%) and neurologic drugs (29.19%). The DiAna dictionary, as a dynamic open-source tool, provides transparency and flexibility, enabling researchers to actively shape drug definitions during the mapping phase. This empowerment enhances accuracy, reproducibility, and interpretability of results.

Enhancing Model Performance through Translation-based Data Augmentation in the context of Fake News Detection

The rapid development of social media in recent years has encouraged the sharing of vast amounts of data and the propagation of fake news. This has pushed the scientific community to focus on this phenomenon, particularly those working on natural language processing, by developing detection tools to combat fake news. At the same time, most studies have focused on languages with a high resource content (corpus). This paper aims to shed light on low-resource languages, particularly the Algerian dialect, through an experimental study with two objectives. The first one is to verify if the automatic translation from Modern Standard Arabic (MSA) to the Algerian dialect can be considered an approach to increase the resources in the Algerian dialect, especially with the rise of large language models (LLMs). The second is to verify the impact of the translation-based data augmentation method on fake news detection by using transformer-based Arabic pre-trained models in different data augmentation configurations. We have discovered that LLMs can generate translations that closely resemble human translations. In this study, we demonstrate that data augmentation can result in saturation and a decline in model performance due to the introduction of noise and variations in writing styles.

Application and Challenges of Information Theory in English Translation and Cross-cultural Communication

Abstract With the deepening of China's foreign exchange, the importance of translation is becoming more and more prominent, but some scholars and translators still have great misunderstandings about translation, and translation has been in a neglected position. Aiming at this phenomenon, this paper introduces information theory into English translation and cross-cultural communication. It proposes an automatic translation control method based on the information entropy model to improve the accuracy and efficiency of translation. It expands the lexicon based on the information entropy model, which is divided into basic lexicon, specialized lexicon, and neologism lexicon. Then, it constructs a semantic model to complete the automated extraction of lexicon features and realize automatic translation. The transfer of language information and the modification of terminological structure are discussed in cross-cultural communication to improve the challenges of English translation and cross-cultural communication. During the translation process, the correlation between the entropy of Chinese and English corresponding unitary, binary, and ternary words - sentence entropy decreases gradually (0.993->0.977->0.972). The largest value of information entropy loss in different cultural contexts is caused by the tonal-phonetic contrast of combining yinping and declination. The relative degree of information loss is 5.300585%. This study has provided new perspectives for translation research through the use of redundant information theory and discrete information theory.

Research on Optimization Strategies for Accuracy of English Automatic Language Translation under Machine Learning Framework

Abstract The mainstream machine translation model Transformer is completely based on the self-attention mechanism for translation operation, but there are still some problems, such as not being able to combine the syntactic structure information of the natural language for translation, which leads to problems such as mistranslation and omission. In this paper, for the problem that the position encoding obtained by traditional RNN and attention mechanism machine translation models using a fixed formula does not contain contextual information, the source language sequences containing contextual positional information are obtained by introducing a bidirectional long-short-term memory network and a tree-shaped long-short-term memory network, trained horizontally and vertically, respectively, and the self-attention mechanism is used in Tree-LSTM for the prediction of the contribution of the decision that The relative position information between words is preserved to the maximum extent, and finally, the Bi-Tree-LSTM translation model based on positional encoding optimization is constructed. The performance of the model is tested on four datasets: general, legal, business, film, and television, and the BLEU value of the model translation is analyzed under low data resources and increased sentence length, and then a 4000-sentence long English text is translated to check the wrong sentences and analyze the translation quality. It was found that the BLEU values of this paper’s model are 33.5, 35.2, 31.7, and 34.4 in the four types of text tests, which are the highest among the models. The BLEU of this paper’s model at 5K data volume has been as high as 26.14 points, which is 2.72 points higher than the highest score of the rest of the machine translation models at 50K data volume. The BLEU scores for 8-18 word sentences consistently remain above 45, and the peak performance is superior to that of other models. 4000 sentences of English long text translation, the total number of error sentences is 54, accounting for 1.39% of the whole text, which is lower than that of the Transformer model’s 7.15%, and the performance is in line with the expectation of the optimization design. This paper provides a new idea and useful exploration for optimizing English machine translation accuracy.

Automatic Machine Translation Quality Evaluation Model under Feature Extraction Algorithm

Automatic Machine Translation Quality Evaluation Model under Feature Extraction Algorithm