Machine Translation Research Articles

Preservation of sentiment in machine translation of low-resource languages: a case study on Slovak movie subtitles

AbstractThis research explores the effectiveness of machine translation from Slovak to English for sentiment analysis, specifically focusing on the translation of movie subtitles. The study employs a parallel corpus of segmented movie subtitles in both languages and utilizes IBM Watson™ Natural Language Understanding service and Google Translate. The research aims to assess the correlation between human-generated text and machine-translated text in sentiment analysis. A comparative analysis was also conducted using OpenAI to evaluate the sentiment of the Slovak text directly, without translation into English. The findings reveal a strong correlation between human text and machine translation, with a Pearson correlation coefficient of 0.86, and a correlation with OpenAI’s GPT model evaluation at 0.72. Despite the relatively high accuracy of the end-to-end solution using OpenAI, the methodology comprising machine translation followed by sentiment analysis in English was found to be significantly more precise. The research further investigates the challenges in translating specific language nuances, such as humor and vulgarism, and their impact on sentiment analysis. The study concludes that machine translation can be effectively used for sentiment analysis in Slovak, a flective language, and highlights the potential of advanced language models in low-resource languages. Future research directions include expanding the study to other text types and comparable languages beyond Slovak.

APPLICATION OF HYBRID APPROACH FOR WOLAITA LANGUAGE PART OF SPEECH TAGGING

The main purpose of this study is to develop part-of-speech tagger for Wolaita Language using hybrid approach. Part of speech tagger is one of the subtasks in NLP application which is important for other Natural Language Processing (NLP) applications, like parser, machine translator, speech recognizer and search engines. PoST is a process of tagging a corresponding part of speech tag for a word that tag defines how the word is used in a sentence. The PoST for Wolaita language is not enough yet to be used as one vital module in other natural language processing applications. In this study, the development of PoS tagger using hybrid approach that combines HMM and rule based approaches was conducted for Wolaita language. In general HMM model need large data to increase the performance and the rule based model learner rule based on the language features. The HMM tagger, tags the words based on the optimal path for a given sequence of words and transformation based learning (TBL) is a rule based approaches that learns rule directly from the training corpus without expert knowledge. The developed hybrid approach of Wolaita language PoS tagger uses HMM tagger as initial annotators and rule based tagger as a corrector based on fixed threshold values. For implementation and experiment purpose the researcher used python programming and NLTK. For training and testing the models, 1256 sentences or 15,268 words are collected from three different categories (Bible, Social media in Wolaita language (Wogetta FM 96.6 ) and Wolaita language department) and annotated data manually. For tagging purpose 26 PoS tag are identified. From entire corpus, 90% for training and the remaining of entire corpus for testing purpose. The performance of the taggers, are tested by using different experiments. After experiment the researcher found that the performance of HMM, rule based and hybrid taggers shows 88.14%, 93.19% and 94.82% respectively. Generally, hybrid approach showed the better performance to assigning part of speech tag for Wolaita language.

Current Trends in the Use of Machine Learning for Error Correction in Ukrainian Texts

The article's authors have provided a detailed problem description of identifying and correcting errors in Ukrainian-language texts. This paper provides a detailed analysis of the latest research and publications aimed at solving the problems of identifying and correcting errors in Ukrainian-language texts. The analysis of modern tools related to error correction in texts is presented along with a comparative description. Investigated the existing data corpora for the Ukrainian language so that they are relevant to solving GEC tasks. Discovered the need to create a large annotated data corpus, which will be prepared by a special team with linguistic expertise. Analysed the opportunities, advantages and disadvantages of modern machine learning models that interpret the task of detecting and correcting errors in texts as classification or machine translation. Introduced the need to develop a machine-learning algorithm that will take into account the specifics of morphologically complex languages, such as Ukrainian. Demonstrated the work of the modern models and provided screenshots. Revealed the need for further research in the Ukrainian segment of machine learning to solve the problems of correcting errors in texts using various methods and approaches.

Marathi to Indian Sign Language Machine Translation

Machine translation has been a prominent field of research, contributing significantly to human life enhancement. Sign language machine translation, a subfield, focuses on translating spoken language content into sign language and vice versa, thereby facilitating communication between the normal hearing and hard-of-hearing communities, promoting inclusivity. This study presents the development of a ‘sign language machine translation system’ converting simple Marathi sentences into Indian Sign Language (ISL) glosses and animation. Given the low-resource nature of both languages, a phrase-level rule-based approach was employed for the translation. Initial encoding of translation rules relied on basic linguistic knowledge of Marathi and ISL, with subsequent incorporation of rules to address 'simultaneous morphological' features in ISL. These rules were applied during the ‘generation phase’ of translation to dynamically adjust phonological sign parameters, resulting in improved target sentence fluency. The paper provides a detailed description of the system architecture, translation rules, and comprehensive experimentation. Rigorous evaluation efforts were undertaken, encompassing various linguistic features, and the findings are discussed herein. The web-based version of the system serves as an interpreter for brief communications and can support the teaching and learning of sign language and its grammar in schools for hard-of-hearing students.

Unveiling Text Representation with 'Bag of Words'

Techniques for natural language processing (NLP) have grown to be essential tools for deciphering and drawing insightful conclusions from massive volumes of text data. A thorough review of numerous natural language processing (NLP) techniques, including as tokenization, stemming, lemmatization, named entity recognition, sentiment analysis, and topic modelling, is provided in this abstract. These methods are essential for applications like sentiment analysis, machine translation, text assistant categorization, and information retrieval. Furthermore, the capabilities of NLP systems have been greatly improved by recent developments in deep learning, especially with models like BERT and GPT. This has allowed them to reach state-of-the-art performance in a variety of language understanding tasks. The difficulties and potential paths for future study in NLP, including managing ambiguity, comprehending context, and enhancing multilingual assistance, are also highlighted in this abstract. Using NLP tools to their full potential, researchers.

Decoding Complexity: A Mathematical Framework for Enhanced Translation Comprehension

Machine translation tools have demonstrated substantial progress in enhancing translation accuracy since the emergence of artificial intelligence. However, challenges persist in reasoning (or the lack thereof), considering contexts, addressing specific word games, and interpreting very long or very short sentences—those exceeding 50 and falling below 7 words (Bowker, 2023 : 893). Additionally, accurately translating technical or specialized terms and their variations remains a hurdle. This research introduces a categorical mathematical formalization of the comprehension stages in translation, along with a model for calculating acceptances (specific meanings of words) during the verification of meaning hypotheses. The goal is to elucidate the comprehension process and integrate contextual considerations. The formalism delineates a series of fundamental cognitive operations involved in comprehension. Furthermore, it advocates for evaluating meaning hypotheses using logical modalities, particularly hypostases, described as phrases (groups of words)—a unit of discourse rather than language—signifying the structure of arguments conveying the speaker's knowledge. The strength of our proposed mathematical model lies in its independence from both source and target languages, as well as the subjectivity of text authors or translators. Additionally, the assessment of meaning hypotheses relies on verifiable logical modalities, ensuring a reliable, explicable, and controllable outcome.

Bridging Languages through Images: A Multilingual Text-to-Image Synthesis Approach

This research investigates the challenges posed by the predominant focus on English language text-to-image generation (TTI) because of the lack of annotated image caption data in other languages. The resulting inequitable access to TTI technology in non-English-speaking regions motivates the research of multilingual TTI (mTTI) and the potential of neural machine translation (NMT) to facilitate its development. The study presents two main contributions. Firstly, a systematic empirical study employing a multilingual multi-modal encoder evaluates standard cross-lingual NLP methods applied to mTTI, including TRANSLATE TRAIN, TRANSLATE TEST, and ZERO-SHOT TRANSFER. Secondly, a novel parameter-efficient approach called Ensemble Adapter (ENSAD) is introduced, leveraging multilingual text knowledge within the mTTI framework to avoid the language gap and enhance mTTI performance. Additionally, the research addresses challenges associated with transformer-based TTI models, such as slow generation and complexity for high-resolution images. It proposes hierarchical transformers and local parallel autoregressive generation techniques to overcome these limitations. A 6B-parameter transformer pretrained with a cross-modal general language model (CogLM) and fine-tuned for fast super-resolution results in a new text-to-image system, denoted as It, which demonstrates competitive performance compared to the state-of-the-art DALL-E-2. Furthermore, It supports interactive text-guided editing on images, offering a versatile and efficient solution for text-to-image generation.. Keywords: Text-to-image generation, Multilingual TTI (mTTI), Neural machine translation (NMT), Cross-lingual NLP, Ensemble Adapter (ENSAD), Hierarchical transformers, Super- resolution, Transformer-based models, Cross-modal general language model (CogLM).

Multilingual Neural Machine Translation for Indic to Indic Languages

The method of translation from one language to another without human intervention is known as Machine Translation (MT). Multilingual neural machine translation (MNMT) is a technique for MT that builds a single model for multiple languages. It is preferred over other approaches, since it decreases training time and improves translation in low-resource contexts, i.e., for languages that have insufficient corpus. However, good-quality MT models are yet to be built for many scenarios such as for Indic-to-Indic Languages (IL-IL). Hence, this article is an attempt to address and develop the baseline models for low-resource languages i.e., IL-IL (for 11 Indic Languages (ILs)) in a multilingual environment. The models are built on the Samanantar corpus and analyzed on the Flores-200 corpus. All the models are evaluated using standard evaluation metrics i.e., Bilingual Evaluation Understudy (BLEU) score (with the range of 0 to 100). This article examines the effect of the grouping of related languages, namely, East Indo-Aryan (EI), Dravidian (DR), and West Indo-Aryan (WI) on the MNMT model. From the experiments, the results reveal that related language grouping is beneficial for the WI group only while it is detrimental for the EI group and it shows an inconclusive effect on the DR group. The role of pivot-based MNMT models in enhancing translation quality is also investigated in this article. Owing to the presence of large good-quality corpora from English (EN) to ILs, MNMT IL-IL models using EN as a pivot are built and examined. To achieve this, English-Indic Language (EN-IL) models are developed with and without the usage of related languages. Results show that the use of related language grouping is advantageous specifically for EN to ILs. Thus, related language groups are used for the development of pivot MNMT models. It is also observed that the usage of pivot models greatly improves MNMT baselines. Furthermore, the effect of transliteration on ILs is also analyzed in this article. To explore transliteration, the best MNMT models from the previous approaches (in most of cases pivot model using related groups) are determined and built on corpus transliterated from the corresponding scripts to a modified Indian language Transliteration script (ITRANS). The outcome of the experiments indicates that transliteration helps the models built for lexically rich languages, with the best increment of BLEU scores observed in Malayalam (ML) and Tamil (TA), i.e., 6.74 and 4.72, respectively. The BLEU score using transliteration models ranges from 7.03 to 24.29. The best model obtained is the Punjabi (PA)-Hindi (HI) language pair trained on PA-WI transliterated corpus.

Evaluating NMT using the non-inferiority principle

Abstract The aim of this article is to propose a new neural machine translation (NMT) evaluation method based on the non-inferiority principle. In order to do that, we evaluate raw machine translation (MT) in terms of naturalness, which for this research is defined as not just the lack of fluency errors but also meeting the linguistic expectations of Galician end users when reading original texts in Galician. Our main objective is, in the first place, to validate the new methodology presented in our previous study by evaluating an NMT engine from Spanish into Galician for the social media domain that was retrained with a new Twitter corpus. This new methodology and NMT engine were applied after analyzing the conclusions of a pilot survey conducted among Twitter users to evaluate their perception of tweets translated from Spanish into Galician with our NMT engine created with a corpus of tweets. As in our preliminary study, our aim is to propose a robust quality approximation method based on the reception parameters of end users’ perceptions. This new survey was conducted in December of 2022 with the participation of 228 Galician-speaking Twitter users. Among the main changes proposed are the inclusion of more information about the participant profile, so the non-inferiority principle can be also evaluated according to these parameters; the inclusion of a new typology of tweets, the threads; the provision of context by means of presenting the tweets in their original display as shown in the Twitter app; a change in the number of tweets evaluated and the number of different questionnaires; the change in the distribution of the questionnaires; and the inclusion of an error classification human evaluation conducted by professional linguists to correlate the findings. We will present the steps carried out following the conclusions of the pilot study, describe the new study’s design, analyze the new findings, and present the final conclusions regarding the engine and the evaluation method based on the non-inferiority principle. Finally, we will also provide some examples of the use of this new methodology in the translation industry.

Data augmentation and transfer learning for cross-lingual Named Entity Recognition in the biomedical domain

AbstractGiven the increase in production of data for the biomedical field and the unstoppable growth of the internet, the need for Information Extraction (IE) techniques has skyrocketed. Named Entity Recognition (NER) is one of such IE tasks useful for professionals in different areas. There are several settings where biomedical NER is needed, for instance, extraction and analysis of biomedical literature, relation extraction, organisation of biomedical documents, and knowledge-base completion. However, the computational treatment of entities in the biomedical domain has faced a number of challenges including its high cost of annotation, ambiguity, and lack of biomedical NER datasets in languages other than English. These difficulties have hampered data development, affecting both the domain itself and its multilingual coverage. The purpose of this study is to overcome the scarcity of biomedical data for NER in Spanish, for which only two datasets exist, by developing a robust bilingual NER model. Inspired by back-translation, this paper leverages the progress in Neural Machine Translation (NMT) to create a synthetic version of the Colorado Richly Annotated Full-Text (CRAFT) dataset in Spanish. Additionally, a new CRAFT dataset is constructed by replacing 20% of the entities in the original dataset generating a new augmented dataset. We evaluate two training methods: concatenation of datasets and continuous training to assess the transfer learning capabilities of transformers using the newly obtained datasets. The best performing NER system in the development set achieved an F-1 score of 86.39%. The novel methodology proposed in this paper presents the first bilingual NER system and it has the potential to improve applications across under-resourced languages.

Cognitive Load and Translation Accuracy in Technology Assisted Simultaneous Interpreting Enabled by Artificial Intelligence

This paper uses eye-tracking research to examine the cognitive load and translation accuracy of simultaneous interpretering with and without technology-assisted in the direction of Chinese-to-English translation and to explore the moderating role of simultaneous interpreting ability. It was found that the new type of simultaneous interpreting assisted by speech recognition technology and machine translation technology can significantly reduce the cognitive load of student interpreters and improve the accuracy of translation. However, the advantages of technology-assisted simultaneous interpreting are not significant enough for professional interpreters compared with student interpreters. This paper explains the results of the study from the perspectives of memory pressure, bilingual switching pressure, and the “ceiling effect”.

Bilingual Translation in Spiderman the Movie Storybook

The language transfer can be seen in the lexical choices in a translation. A translator may have different considerations when applying its use in a text since adaptation of context is an essential part to be employed through the translation. Problems with transferring the source language into the target language certainly have an effect on delivering messages within a text. This study aims to analyze the lexical usage in the bilingual translation of the movie storybook Spiderman and the strategies in the translation, including their meaning. The qualitative method used DeepL to assist in the analysis of differences that might come into the translation. In some cases, words translated from English into Indonesian were found non-equivalent, however, the rest of them were understandable. Translation by a more general word strategies were also used to be placed in the story rather than using the specific words and vice versa, paraphrasing using a related word, omission, and neutral words or less expressive words found to correspond the meaning. The equivalence of transferring the source language (SL) into the target language (TL) can be achieved when optional words are found in the translation in which meaning is also taken for consideration within the human translation. The product of translation was presented bilingually to analyze the differences in meaning between the two languages’ context of use. These showed DeepL as one of the machine translations with translation memories using Artificial Intelligence technology which compares the differences and similarities of sense in translation study. They can work together for the analysis to understand the use and results. Keywords: equivalence, non-equivalence, lexical, strategy

Examining the Cultural Connotations in Human and Machine Translations: A Corpus Study of Naguib Mahfouz's Zuqāq al-Midaqq

The translation of culture-specific terms constitutes a major challenge for professional translators as it necessitates a thorough understanding of both the linguistic and cultural elements. With rapid technological advancement over the past few years, machine translation has enhanced translation quality. This study investigates the transability of cultural connotations in the Arabic-English translation of Naguib Mahfouz’s Zuqāq al-Midaqq. A descriptive qualitative research design was adopted to achieve the intended goals of the study. The data comprised human and machine translations from Google Translate and ChatGPT. Through qualitative content analysis, the translations were compared for accuracy in transferring the cultural connotations prevalent in the Arabic source text. The findings revealed that the human translation showed greater naturalness and accuracy in rendering cultural connotations. Machine translation has struggled with rhetorical devices, idioms, and cultural nuances. The results also indicated that the AI-enhanced machine represented by ChatGPT captured the cultural elements more effectively than Google Translate. The study concluded that human expertise remains essential for the high-quality translation of literary works to maintain cultural significance. The findings can inform translator training and guide improvements to AI-enhanced translation for literary texts.

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.

Machine Translation in EFL Learning: Chinese Undergraduates’ Use, Perceptions, and Machine Translation Literacy

Machine translation (MT) has been recently experiencing an upsurge in popularity within the realm of second language (L2) and foreign language (FL) education.This paper aims to present the findings of a questionnaire survey that investigated the use and perception of MT, as well as MT literacy, in English as a foreign language (EFL) learning among undergraduate students in China. A total of 936 online questionnaires were collected from one university in China, with corresponding statistical data analyzed thereafter. The results indicate that MT is prevalently and frequently used in EFL learning by both low- and high-proficiency respondents from diverse disciplines, highlighting an urgent need to improve their MT literacy. This study holds significant pedagogical implications, shedding light on best practices for the efficient use of MT in realm of language education, particularly in EFL education.

Reading against all odds: creative coping strategies of international students in HE

ABSTRACT Many overseas students in Higher Education in the UK struggle to understand the compulsory texts for their course, and obtain lower scores for their modules than their monolingual peers. While the existence of this achievement gap is well established in the literature, little is known about the ways in which overseas students in HE with limited English language and literacy skills meet the challenges posed by the compulsory reading on their course. The current study is novel because we focus on a postgraduate Law module and use a mixed methods approach to study the relationship between students’ language and literacy profiles and their academic achievement. The results show that students’ vocabulary sizes did not provide enough lexical coverage for them to understand the set texts and that the language and literacy profiles of our students did not correlate with their module scores. In-depth interviews revealed that students with IELTS scores of 6.0 or lower used ‘creative reading’ strategies such as machine translation, to pass assignments, while staff used ‘compassionate marking’ strategies in the case of struggling writers, which means module scores for some students were likely inflated. Implications for admissions tutors and teachers in HE are formulated at the end.

A quantitative evaluation of online translators using Hindi web queries

Understanding machine translation (MT) quality is becoming more and more important as MT usage continues to rise in the translation industry. The acceptance of MT output based on their performance and, ultimately, how acceptable the translators actually are, have received relatively less attention so far. MT plays a vital role in CLIR systems and their retrieval efficiency is directly proportional to the translation accuracy of the queries. The varied meanings of words, sentences carrying multiple interpretations, and differing grammatical structures across languages contribute to the complexity of the MT task. The lack of structural constraints and the presence of ambiguity further compound the complications especially in case of web queries. The objective of this work is to assess the accuracy of free online translators in translating Hindi web queries. The accuracy of the translators has been evaluated on various metrics, i.e., BLEU, NIST, METEOR, hLepor, CHRF and GLEU. Our findings indicate that the translation accuracy for longer queries is higher than the shorter ones. Overall Google translator’s performance has been found the best while Systran performs the worst with 42.06% performance difference between the two. The present work intends to help researchers in further evaluating and analyzing the MT systems specially in context of web query translation, ultimately leading to improved translation quality and retrieval accuracy in CLIR.

SOFTWARE CAPABILITIES FOR TEXT IDENTIFICATION: COMPARISON FOR SIMILARITY, IDENTITY ESTABLISHMENT, UNIQUENESS CHECK

The study was carried out within the framework of identification linguistics and translational linguistics. The article describes some methods for determining the degree of similarity of texts: shingle algorithm, Levenshtein distance, systems for detecting plagiarism. The purpose of the work is to test the software capabilities of comparing texts for similarity, establishing their identity, and checking uniqueness. In a broad sense, these tasks fall within the area of text identification. In a qualitative (manual) assessment of the similarity of texts, identifying parameters are selected and selected specifically for the text under study. The use of electronic resources is determined by the desire for objectivity of the methods used to establish the identity of texts and the objectivity of the results obtained. Software products also make it possible to establish another, quantitative, characteristic — the degree of similarity of texts to each other or the degree of originality of the text. The work used services whose tasks include 1) comparing the similarity of two texts; 2) calculation of the Levenshtein distance; 3) detection of borrowing. The research material was an excerpt from an interview with Foreign Minister Sergei Lavrov. Reverse machine translation texts served as options for comparison with the source text. Reverse machine translation as a translation product is part of artificial intelligence and a model of the process of understanding and interpreting natural language. The results of using the proposed services made it possible to arrange five reverse machine translation options from the most unique text to the text that is most identical to the invariant. The study showed that the programs generally produce similar results, which can be applicable to solving research and applied problems related to establishing the identity and difference of texts. The prospect of the study is to identify lexical parameters that make it possible to classify reverse machine translation texts as the most or least identical with respect to the invariant.

Application of ELAN in Human-Machine Mutual Translation of English for the News

The translation of English news means the process of converting news articles published in English into another language. These articles may also be referred to as English news. This process comprises understanding the substance of the source article and accurately translating its tone and intent into the target language. A human translator or machine translation software can be used to perform the translation manually. In the disciplines of software globalization and technological translation, the Human-Machine Mutual Translation technique has grown significantly with the advancement of civilization and the quickening growth of research and innovation. Because of this practical need, translators have begun to focus more on machine translation and have conducted some productive research in this area. As a result, we suggested the EUDICO Linguistic Annotator (ELAN) in Human-Machine mutual translation of English for the news. EUDICO stands for European Distributed Corpora Project. The Max Planck Institute for Psycholinguistics created the EUDICO Linguistic Annotator as a software program for the linguistic annotation of language data. For this study, the English news dataset was used. The data were pre-processed by using normalization. The results were evaluated using performance metrics such as translation quality, precision, BLEU, recall, and computation time. The findings show that the recommended procedure provides an accurate translation of English news when compared to previously used methods. The results demonstrated that translation quality and computation time of proposed model are 95% and 60s respectively.

Sanskrit to Hindi language translation using multimodal neural machine translation

Machine translation (MT) is a subfield of computer features that focuses on the automatic translation from one natural language into another without any human involvement. Due to native people interacting in a variety of languages, there is a great need for translating information between languages to send and communicate thoughts. However, they disregard the significance of semantic data encoded in the text features. In this paper, multimodal neural machine translation (MNMT) is proposed for Sanskrit-Hindi translation. The main goal of the proposed method is to fully utilize semantic text features on NMT architecture and to minimize testing and training time. The MNMT is validated on two different NMT architectures: recurrent neural network (RNN) and self-attention network (SAN). The MNMT method’s efficacy is demonstrated by employing the dataset of Sanskrit-Hindi Corpora. Extensive experimental outcomes represent the proposed method’s enhancement over baselines on both architectures. The existing methods, namely, English-to-Indian MT system, Sanskrit-Hindi MT system, and hybrid MT system are used to justify the efficacy of the MNMT method. When compared to the above-mentioned existing methods, RA-RNN respectively achieves a superior BLEU and METEOR of 80.5% and 75.3%, while the RA-SAN respectively achieves a superior BLEU and METEOR of 78.2% and 77.1%.