MT System Research Articles

Language Translator Using Deep Learning

Abstract: Digital translation (MT) is a process of translating text from one language to another using software by including both computer and language information. Initially, the MT system acquires text translation into the source language by simply associating the meaning of the words in the source language with the target language with the help of grammar. However, such methods did not produce good results due to their failure to capture the various sentence structures in the language. This process of translation is time-consuming and requires skilled craftsmen in both languages. Subsequently, integrated translation methods such as mathematical translation (SMT) and neural translation (NMT) technology have been introduced to address the challenges of legal-based approaches. MT has already shown promising results in bilingual translation. In contrast to SMT, which requires sub-components trained separately in translation, NMT uses one large neural network for training. This paper outlines how to train a repetitive neural network for rearrangement for a source to identify. The default encoder helps to reconstruct the vectors of the target language. Therefore, powerful hardware (GPU) support is required. The GPU improves system performance by reducing training time.

The use of residual analysis to improve the error rate accuracy of machine translation

The aim of the study is to compare two different approaches to machine translation—statistical and neural—using automatic MT metrics of error rate and residuals. We examined four available online MT systems (statistical Google Translate, neural Google Translate, and two European commission’s MT tools—statistical mt@ec and neural eTranslation) through their products (MT outputs). We propose using residual analysis to improve the accuracy of machine translation error rate. Residuals represent a new approach to comparing the quality of statistical and neural MT outputs. The study provides new insights into evaluating machine translation quality from English and German into Slovak through automatic error rate metrics. In the category of prediction and syntactic-semantic correlativeness, statistical MT showed a significantly higher error rate than neural MT. Conversely, in the category of lexical semantics, neural MT showed a significantly higher error rate than statistical MT. The results indicate that relying solely on the reference when determining MT quality is insufficient. However, when combined with residuals, it offers a more objective view of MT quality and facilitates the comparison of statistical MT and neural MT.

Word Sense Disambiguation applied to Assamese-Hindi Bilingual Statistical Machine Translation

Word Sense Disambiguation (WSD) is concerned with automatically assigning the appropriate sense to an ambiguous word. WSD is an important task and plays a crucial role in many Natural Language Processing (NLP) applications. A Statistical Machine Translation (SMT) system translates a source into a target language based on phrase-based statistical translation. MT plays a crucial role in a WSD system, as a source language word may be associated with multiple translations in the target language. This study aims to apply WSD to the input of the MT system to enhance the disambiguation output. Hindi WordNet was used by selecting the most frequent synonym to obtain the most accurate translation. This study also compared Naïve Bayes (NB) and Decision Tree (DT) to test and build a WSD model. NB was more appropriate for the WSD task than DT when evaluated in the Weka machine learning toolkit. To the best of our knowledge, no such work has been carried out yet for the Assamese Indo-Aryan language. The applied WSD achieved better results than the baseline MT system without embedding the WSD module. The results were analyzed by linguist scholars. Furthermore, the Assamese-Hindi transliteration system was merged with the baseline MT system for the translation of proper nouns. This study marks a remarkable contribution to Assamese NLP, which is a low computationally aware Indian language.

Exploring Translation Strategies of Japanese Manga in Google Translate and Komikcast Translations

This paper examines the translations of ‘Golden Kamuy Vol. 1’ from Japanese to Bahasa Melayu/Bahasa Indonesia by MT system Google Translate (GT) and online fan-based translating community Komikcast. Given the linguistic and sociocultural differences between the source language (SL) and target language (TL), translating manga demands expertise and a profound understanding of the cultural nuances embedded in the SL, the specific manga genre, and the target audience. This qualitative descriptive study uses Vinay and Darbelnet’s (1958/1995) translation strategies to analyze the translations of ‘Golden Kamuy Vol. 1’ by GT and Komikcast. The study finds that both platforms employ Literal Translation, Borrowing, and Equivalence strategies, with Komikcast additionally using Modulation. GT primarily relies on literary translation, which is consistent with its algorithmic nature and reliance on translation memories. Conversely, Komikcast emphasizes the Equivalence strategy, recognizing the significance of language proficiency and the contextual and cultural insights human translators provide. Findings are discussed in relation to Nida’s equivalence effects, particularly regarding the translated text’s purpose, translators’ knowledge, and cultural differences in SL and TL. This research enhances our understanding of manga translation, as well as the strengths and limitations of machine and human translation, while contributing to the broader discussion on translation studies. While acknowledging the role and potential of machine translation, this research also reinforces the critical role of human translators in bridging linguistic and cultural gaps between Japanese and Indonesian, highlighting the need for translators to possess deep cultural insights and proficiency beyond literal translations for more authentic and contextually accurate renditions.

Effectiveness of Machine Translation in Rendering Yemeni Culture-Specific Items into English: Sana'ani Dialect as a Case-in-Point

Undoubtedly, in any given society cultural elements are manifested in all aspects of life. In this globalized age, efforts have been exerted in translating cultural elements. However, since each culture has its own linguistic peculiarities and social aspects, it has become a complicated issue in the field of MT. Despite the recent technological developments, MT still lacks the ability to produce adequate translation in terms of author-based accuracy and reader-oriented cultural acceptability. This study aims to assess the effectiveness of MT in rendering Yemeni culture-specific items into English by comparing the translation outcomes of three MT programs (Google Translate, Yandex Translate and Bing Translator) to find out which program can produce the closest translation to the source text. Moreover, the study aims to identify translation problems of Yemeni culture-specific items resulting from MT. To achieve these aims, the researchers followed a descriptive qualitative method within a comparative framework. The study sample includes 14 Yemeni culture-specific items in their contexts taken from the book Social Issues in Popular Yemeni Culture by Abdu Al Rahman Mutahar and inserted into the three MT systems. The study concluded that MT lacks the ability to translate Yemeni culture-specific items adequately. The three programs failed to render most of the selected sample adequately. Google and Yandex failed to translate 13 cultural items whereas Bing made 12 errors in translating the same cultural items. The study results also showed that MT faced challenges in translating Yemeni culture-specific items, resulting in lexical, semantic and pragmatic problems.

Comparing the Performance of Google Translate and SYSTRAN on Arabic Lexical Ambiguity

Machine translation systems face many challenges when dealing with Arabic lexical ambiguity, which affects the quality and accuracy of their translations. This study investigates how two popular MT systems, Google Translate and SYSTRAN, handle three problematic linguistic features of the Arabic language: homonyms, heteronyms, and polysemes. A test suite was designed to include sentences that contain these features in different contexts and domains. The translations produced by the two MT systems were evaluated by four independent evaluators for intelligibility and accuracy using a four-point scale. Results showed that both MT systems struggled with the three chosen linguistic features, with average scores below 40%. Google Translate outperformed SYSTRAN in almost every sentence in the test suite. Both systems scored better in intelligibility than in accuracy. Heteronyms proved to be the most challenging for both MT systems due to the unique design of Arabic discretization, which is not yet recognized by translation systems. This study contributes to the field of machine translation by providing a comprehensive analysis of Arabic lexical ambiguity and its impact on MT quality, as well as suggesting possible ways to improve MT systems for Arabic-English translation.

Machine versus corpus-based translation of multiword terms

Abstract Machine translation (MT) post-editing is an increasingly common practice in the translation industry which is also slowly being applied in the development of terminological resources. However, more studies have been devoted to analyze the practice in a translation scenario than in a terminographic context. Consequently, term-oriented post-editing guidelines are a current need if terminographers are also to become post-editors. With a view to enhancing the multilingual representation of environmental multiword terms (MWTs) in terminological resources, we analyze English–Spanish MWT translation in various generic MT systems. Our aims are: (1) to evaluate MT output in order to check whether it can be of any help to terminographers’ work; (2) to develop an error typology in order to raise terminographers’ awareness; and (3) to use the error typology to sketch a series of basic pre-editing and post-editing rules in a terminographic scenario. A comparison of MT output with the equivalents found in a comparable corpus is also presented. Even though MT often presents errors or unidiomatic choices, it can still serve as a basis for human post-editing, and provided that post-editors are familiarized with the potential errors. Comparable corpora, on the other hand, offer better results, but searches are more time-consuming and equivalents are not always available.

Book Review of Zakaryia Almahasees’s Analysing English-Arabic Machine Translation

Analysing English-Arabic Machine Translation evaluates three prominent MT systems, including Google Translate, Microsoft Translator, and Sakhr, each of which provides translation between English and Arabic in a large corpus and paves the way for further research on such an important topic

Assessing MT with measures of PE effort

Recent improvements in quality obtained by neural machine translation (NMT) have boosted its presence in the translation industry. In many domains and language combinations, translators post-edit raw MT output: they edit and correct the pre-translated text to produce the final translation. However, this process can only produce the expected results if the quality of the raw MT can be assured. MT is usually assessed with automatic metrics, as they are faster and cheaper. However, these metrics are not always good quality indicators and do not correlate to the post-editing effort.We suggest a two-step evaluation process for MT intended for post-editing. The automatic evaluations are followed by the assessment of the three dimensions of PE effort. This targeted evaluation can ensure a quality of the raw MT which does not jeopardise the final product or compromise the task of post-editors. We include a detailed description of PosEdiOn, an easy-to-use standalone tool which records PE effort, and a use case of its implementation. 18 translators post-edit texts from English into Spanish from the news domain translated with DeepL and an NMT system trained by the authors to gather PE effort metrics. We compare automatic and PE effort metrics to assess which MT system would be more suitable for post-editing.

Do cues in a video help in handling rare words in a machine translation system under a low-resource setting?

Video modality is an emerging research area among the numerous modalities utilized for multimodal machine translation. Multimodal machine translation uses multiple modalities to improve the machine-translated target language from the source language. However, the currently available multimodal dataset is focused on a few well-studied languages. In this paper, we propose a video-guided multimodal machine translation (VMMT) model under a low-resource setting by building a synthetic multimodal dataset of the English-Hindi language pair, the first one of its kind for this language pair. The VMMT system employs spatio-temporal video context as an additional input modality along with the source text. The spatio-temporal video context is extracted using a pre-trained 3D convolutional neural network. We report how well the VMMT systems outperform the text-only neural machine translation (NMT) system using automatic evaluation metrics and human evaluation on two test datasets: one in-domain and another out-domain. Our results indicate that the use of video context as an additional input modality enhances the performance of the MT system in resolving the various MT challenges, such as handling rare words, ambiguity, etc., in both English→Hindi and Hindi→English translations. Our experimental results show a significant improvement of up to +4.2 BLEU and +0.07 chrF scores in English→Hindi and +5.4 BLEU and +0.07 chrF scores in Hindi→English with our VMMT system over unimodal NMT system. Our findings highlight the potential of visual cues as an additional modality for improving machine translation systems especially in low-resource settings and emphasize the importance of synthetic multimodal datasets in addressing the scarcity of diverse data for less-studied language pairs.

Transmission of Power and Signal Multiplex for Electric Vehicles Using Cascaded Multilevel Inverters

Abstract: P&SMT, or power and signal multiplex transmission, is a method for transmitting communication signals using power electronic circuits. A three-phase cascaded multilevel inverter-based P&S MT system is recommended in this investigation. By allowing communication signals to be transmitted without utilizing a Controller Area Network bus, the proposed solution can lower the wiring expense of the existing electric vehicle (EV) communication system. The system's architecture enables battery balancing discharge and motor speed management for EVs. In a simulation system proposed in MATLAB/Simulink, both power and communication signals are effectively conveyed using the combined pulse width modulation scheme and frequency shift keying approach. The maximum signal rate of the proposed system is established at 600 bits per second by analyzing the bit error rate of the sent signal.

Investigating implicature in the machine translation of selected conversations from the novel ‘The hunger games’

This study aims at investigating implicatures in the machine translation of selected conversations from the novel “The Hunger Games” by Collins (2008). The selected conversations are nine conversations taken from the beginning, middle and end of the novel from crucial points in the novel’s plot and are conversations that push the action of the novel forward. This study analyzes conversational implicatures according to Grice’s maxims (1991). It specifically examines the MT system’s translation of instances where these maxims were flouted in the source text. This is done through putting the novel’s conversations in to Google translate and the output translation is then analyzed and compared with the source text’s implicature . The study reaches the following findings: the MT system can translate conversational implied meanings correctly when the maxims of relation or quantity were flouted. This is due to the fact that it translates the utterance literally and thus the same effect is mimicked in the target text. However, the implied meanings created through the flouting of quality or manner maxims are harder for the machine to translate. Therefore, researchers in the field of machine translation should work to improve the machine translation regarding implicature created through the flouting of these two maxims.

Effect of Locally Adapted Conservation Tillage on Runoff, Soil Erosion, and Agronomic Performance in Semiarid Rain-Fed Farming in Ethiopia

An on-farm field experiment on a locally adapted conservation tillage method was undertaken to evaluate its effect on soil erosion, surface runoff, and agronomic parameters. It was conducted on five farmer fields with 3–14% slopes in the Rift Valley and the Eastern escarpment of Ethiopia’s central highlands region for two cropping seasons. The treatments were conventional tillage (CT), repeated ploughing performed with a traditional ox-drawn plough named ‘Maresha’, and minimized contour ploughing (MT) at most twice with a locally adapted sweep-like attachment assembled to Maresha. Surface runoff and soil loss in the MT system were 30 to 60% and 49 to 76% lower than those in the CT system on 3 to 14% slopes, respectively. Despite the wide variation in surface runoff, limited differences in soil water content for the depth from 0 to 20 cm were observed between the treatments. Significant differences (p < 0.05) in grain yields (kg ha−1) of 246 and 323 in the 1st and 2nd growing seasons, respectively, were recorded between the MT and CT treatments. The results of this study demonstrated that the MT system can significantly reduce surface runoff and soil loss while improving crop yields in rainfed smallholder farming systems of Ethiopia.

A Thematic-Role-Based Approach for Word Sense Disambiguation

The present paper investigates the thematic roles that can be developed for the purpose of Word Sense Disambiguation. In MT systems and electronic word databases, thematic role relations are not clearly included among other semantic relations. Identifying thematic roles of predicates helps in disambiguating word senses and hence producing more accurate translation. For instance, the meaning of the verb ‘eat’ differs depending on the thematic roles it assigns for its Subject and Object. When it assigns an animate Agent for its Subject and food Patient for its Object, it means ‘take in solid food’. However, when it assigns Force for its Subject and metal Theme for its Object, it means ‘cause to deteriorate due to the action of water, air or an acid’. Accordingly, different translations are produced in each context. Selectional restrictions are also tackled in the analysis of the sample verbs. The implementation is made on three MT systems: Al Wafi, Sakhr and Google. They all produce incorrect translations of the sample verbs. A suggested translation is proposed for each verb after analyzing its thematic roles and selectional restrictions. In this way, the present paper is significant since it helps in improving the performance of MT systems. The present paper will focus only on a group of English verbs that convey a variety of meanings. It will show a number of problematic cases in translation that occur due to the lack of thematic roles in the core of the system. After developing thematic roles, it is expected that such cases will be disambiguated.

Proposal of High Accuracy and Rapid Screening Method for 2nd Life Battery using Multi-layer AI Analysis and MT System on Taguchi Method

Proposal of High Accuracy and Rapid Screening Method for 2nd Life Battery using Multi-layer AI Analysis and MT System on Taguchi Method

Evaluation of condition on replacing repaired concrete based on NDT and the Mahalanobis–Taguchi system

In Japan, much infrastructure has been constructed over the past 50 years, and maintenance of such infrastructure is important. Repairing and retrofitting of deteriorated concrete members has become a common place. Meanwhile, re-deterioration of repaired concrete members has become of concern. To prevent re-deterioration of the repaired concrete members, it is important to evaluate the quality and condition of repair, in addition to the use of NDE. Therefore, to evaluate the condition of replacing repaired concrete specimens, an ultrasonic test, impact-echo test, and impact test for measuring mechanical impedance were applied. In addition, data of these NDT are integrated by the Mahalanobis–Taguchi system (MT system). The MT system is an information processing system for pattern recognition advocated by quality engineering. From the results, it is clarified that the Mahalanobis distance calculated by the MT method could successfully evaluate the conditions of replacing repaired specimens.

Hindi to English Multimodal Machine Translation on News Dataset in Low Resource Setting

This work proposes a multimodal Hindi-to-English machine translation on a news corpus by integrating multiple input modalities. The experimental dataset comprises of an image from a news article and its caption in the English-Hindi language. The parallel English-Hindi dataset is constructed by translating the image caption from English-language news articles into Hindi. Unlike standard images used for NLP tasks, images in news articles frequently depict significant events, locations, or individuals. The caption provides a more extensive description of the image, resulting in the inclusion of multiple identified entities. Our methodology consists of encoders for each input modality, allowing for simultaneous consideration of both text and image. We undertake two tests to examine how effectively the image content improves MT systems when 1) the entire image is considered first and 2) the largest object recognized in the image is considered. We employ Byte Pair Encoding (BPE) to represent the text, and the visual attributes are extracted from the image with VGG-19, a pre-trained CNN model. In terms of BLEU and chrF, our Multimodal Machine Translation (MMT) systems are superior than the baseline unimodal NMT system. Our MMT system outperforms the NMT model by +1.8 BLEU and +0.03 chrF.

Traitement de l’ambiguïté syntaxique et sémantique en TA neuronale : analyse de la traduction de l’anglais vers le français, l’espagnol et l’italien

Syntactic and Semantic Ambiguity Processing in Neural MT from English to French, Spanish and Italian
 Despite recent advances in artificial intelligence, human translators outperform MT for at least three types of tasks: identifying referents in anaphora (especially of the interphrastic kind), resolving semantic ambiguity (which is mainly due to polysemy or homonymy), and resolving syntactic ambiguity (especially with poorly inflected source languages such as English). Using the results obtained by two freely available online machine translation programs, Google Translate and DeepL, we examine how these two types of ambiguity are processed in translation from English into French, Spanish and Italian. Our results show that the two programs perform well overall in resolving the simplest cases of syntactic ambiguity, with difficulties arising more frequently for noun phrases featuring atypical syntactic divisions and rarely used collocations. MT output for ambiguous structures involving verb roots followed by the –ING morpheme (flying planes, growing pains) is studied, as well as syntactic structures in which two or more nouns are preceded by one or more adjectives. MT handles relatively well the longest of those structures (ADJ ADJ N N N N), probably because their subsets are part of the bilingual or target language monolingual corpora that underlie MT systems. Structures involving head modification and coordination (ADJ N AND N) are also known to pose problems for MT and human translators alike. But since many of the most frequent N AND N structures involve cohyponyms (men and women, brothers and sisters), antonyms (rights and duties, costs and benefits) or near- synonyms (aid and advice), their translation as a whole unit generally triggers the choice of correct syntact dependencies in translation. Structures in which the adjective only modifies the first noun (fresh air and exercise, social sciences and humanities) are much less frequent and are also probably translated as a whole unit. Structures involving premodification, coordination and post-modification may give rise to four distinct types of structures depending on whether long-range dependencies apply (detailed [knowledge and understanding] of the IT industry, [ethnic group] and [place of birth], invaluable [context and [source of information], [close friend] and confidant] of Mr Jones. Structures in which both long-range dependencies apply (integrated prevention and control of pollution) are the ones which most frequently cause errors for MT. Semantic ambiguity cases have been processed with increasing success by MT, especially when collocates vary widely for the main two meanings of homonyms (a well-known example is the word pen). Processing polysemy (for instance the medical use of conditions in pre-existing conditions) is a bit more of a challenge for MT. Other cases involving concentration of several polysemic terms in the same sentence (Changing the placement of beams relative to the staff involves changing the direction of the stems in the beam) also create difficulties for MT when the polysemic terms are used without any of their usual collocates (here in the specialised field of musical edition). Homonymy cases involving grammatical category changes (N-to-V or V-to-N conversion) seem to continue to pose the most difficulties to neural MT, despite increasing consideration of intra- and extraphrastic context. Potentially ambiguous word sequences (treatment increase in as the daily dose and duration of treatment increase), which were processed incorrectly before neuronal MT, are now correctly translated. But word sequences in which one word belongs to a part of speech which is not the most commonly used one (e.g. the noun remains in what remains can be considered) may cause occasional errors. Several examples that involve the verb founder are studied, and they frequently trigger translation of the noun in all three Romance languages (or translations of the verbs find or found due to incorrect segmentation).

Empirical Analysis of Parallel Corpora and In-Depth Analysis Using LIWC

The machine translation system aims to translate source language into target language. Recent studies on MT systems mainly focus on neural machine translation. One factor that significantly affects the performance of NMT is the availability of high-quality parallel corpora. However, high-quality parallel corpora concerning Korean are relatively scarce compared to those associated with other high-resource languages, such as German or Italian. To address this problem, AI Hub recently released seven types of parallel corpora for Korean. In this study, we conduct an in-depth verification of the quality of corresponding parallel corpora through Linguistic Inquiry and Word Count (LIWC) and several relevant experiments. LIWC is a word-counting software program that can analyze corpora in multiple ways and extract linguistic features as a dictionary base. To the best of our knowledge, this study is the first to use LIWC to analyze parallel corpora in the field of NMT. Our findings suggest the direction of further research toward obtaining the improved quality parallel corpora through our correlation analysis in LIWC and NMT performance.

Investigating Contextual Influence in Document-Level Translation

Current state-of-the-art neural machine translation (NMT) architectures usually do not take document-level context into account. However, the document-level context of a source sentence to be translated could encode valuable information to guide the MT model to generate a better translation. In recent times, MT researchers have turned their focus to this line of MT research. As an example, hierarchical attention network (HAN) models use document-level context for translation prediction. In this work, we studied translations produced by the HAN-based MT systems. We examined how contextual information improves translation in document-level NMT. More specifically, we investigated why context-aware models such as HAN perform better than vanilla baseline NMT systems that do not take context into account. We considered Hindi-to-English, Spanish-to-English and Chinese-to-English for our investigation. We experimented with the formation of conditional context (i.e., neighbouring sentences) of the source sentences to be translated in HAN to predict their target translations. Interestingly, we observed that the quality of the target translations of specific source sentences highly relates to the context in which the source sentences appear. Based on their sensitivity to context, we classify our test set sentences into three categories, i.e., context-sensitive, context-insensitive and normal. We believe that this categorization may change the way in which context is utilized in document-level translation.