Some ancient philosophers compared the world to a big animal. This was vigorously opposed by modernity – the Enlightenment and the nineteenth century –, which compared it to a machine. Today, nobo...

Technology has advanced significantly over the past decades. Significant changes have occurred in the field of translation with the development of programs such as Google.translate and Yandex.translator. The presented applications are already being actively implemented in translation agencies to optimize translation activities, where written translations of documents, articles, annotations, etc. must be provided to customers as quick as possible. While working with popular science text, online programs help translators gain time, but this requires to edit the text. (...) The artistic style requires more concentration and dedication, because. the means of expression presented in it require taking into account the context and nuances of the use of certain units of language. Machine translation has the potential to become an indispensable assistant in the hands of a translator. This article discusses machine translation of expressive means, namely metaphors. The study is illustrated with examples and a comparative analysis of the translation of metaphorical units is carried out, the classification of metaphors is identified and an analysis is carried out with which means of expression the translator is able to cope with. The difficulties of translating metaphors are analyzed. (shrink)

Based on neural machine translation, this article introduced the ConvS2S system and transformer system, designed a semantic sharing combined transformer system to improve translation quality, and compared the three systems on the NIST dataset. The results showed that the operation speed of the semantic sharing combined transformer system was the highest, reaching 3934.27 words per second; the BLEU value of the ConvS2S system was the smallest, followed by the transformer system and the semantic sharing combined transformer system. (...) Taking NIST08 as an example, the BLEU values of the designed system were 4.74 and 1.49 higher than the other two systems. The analysis of examples showed that the semantic sharing combined transformer had higher translation quality. The experimental results show that the proposed system is reliable in English content translation and can be further promoted and applied in practice. (shrink)

Introduction of deep neural networks to the machine translation research ameliorated conventional machine translation systems in multiple ways, specifically in terms of translation quality. The ability of deep neural networks to learn a sensible representation of words is one of the major reasons for this improvement. Despite machine translation using deep neural architecture is showing state-of-the-art results in translating European languages, we cannot directly apply these algorithms in Indian languages mainly because of two (...) reasons: unavailability of the good corpus and Indian languages are morphologically rich. In this paper, we propose a neural machine translation (NMT) system for four language pairs: English–Malayalam, English–Hindi, English–Tamil, and English–Punjabi. We also collected sentences from different sources and cleaned them to make four parallel corpora for each of the language pairs, and then used them to model the translation system. The encoder network in the NMT architecture was designed with long short-term memory (LSTM) networks and bi-directional recurrent neural networks (Bi-RNN). Evaluation of the obtained models was performed both automatically and manually. For automatic evaluation, the bilingual evaluation understudy (BLEU) score was used, and for manual evaluation, three metrics such as adequacy, fluency, and overall ranking were used. Analysis of the results showed the presence of lengthy sentences in English–Malayalam, and the English–Hindi corpus affected the translation. Attention mechanism was employed with a view to addressing the problem of translating lengthy sentences (sentences contain more than 50 words), and the system was able to perceive long-term contexts in the sentences. (shrink)

Technology has advanced significantly over the past decades. Significant changes have occurred in the field of translation with the development of programs such as Google.translate and Yandex.translator. The presented applications are already being actively implemented in translation agencies to optimize translation activities, where written translations of documents, articles, annotations, etc. must be provided to customers as quick as possible. While working with popular science text, online programs help translators gain time, but this requires to edit the text. (...) The artistic style requires more concentration and dedication, because. the means of expression presented in it require taking into account the context and nuances of the use of certain units of language. Machine translation has the potential to become an indispensable assistant in the hands of a translator. This article discusses machine translation of expressive means, namely metaphors. The study is illustrated with examples and a comparative analysis of the translation of metaphorical units is carried out, the classification of metaphors is identified and an analysis is carried out with which means of expression the translator is able to cope with. The difficulties of translating metaphors are analyzed. (shrink)

Automated translation is systematically gaining popularity among professional translators, who claim that editing MT output requires less time and effort than translating from scratch. MT technology is also offered in leading translator’s workstations, e.g., SDL Trados Studio, memoQ, Déjà Vu and Wordfast. Therefore, the dilemma arises: should MT be introduced into formal translation training? In order to answer this question, first, it is necessary to understand how trainee translators actually use MT. This study is an attempt to obtain (...) this knowledge. The methodology applied in this investigation is text analysis. During the experiment sessions the students were asked to translate a legal text using MT tools, which in practice meant the post-editing of the MT raw output. The post-edited versions of the text underwent analysis in order to answer the following research questions: - What are the most typical errors contained in both French and English MT output? - How critical are the students towards the text generated by MT? - How perceptive are the students during the post-editing task? - Are they able to detect and correct errors using their knowledge and skills? The results of this study suggest that the post-editing of the MT raw output is as demanding for translation students as traditional translation, however, it requires a different set of skills, such as critical thinking and perceptiveness. Therefore, a special kind of training related to the effective use of MT technology should be implemented during translation classes. (shrink)

In order to improve the efficiency of the English translation, machine translation is gradually and widely used. This study briefly introduces the neural network algorithm for speech recognition. Long short-term memory (LSTM), instead of traditional recurrent neural network (RNN), was used as the encoding algorithm for the encoder, and RNN as the decoding algorithm for the decoder. Then, simulation experiments were carried out on the machine translation algorithm, and it was compared with two other (...) class='Hi'>machine translation algorithms. The results showed that the back-propagation (BP) neural network had a lower word error rate and spent less recognition time than artificial recognition in recognizing the speech; the LSTM–RNN algorithm had a lower word error rate than BP–RNN and RNN–RNN algorithms in recognizing the test samples. In the actual speech translation test, as the length of speech increased, the LSTM–RNN algorithm had the least changes in the translation score and word error rate, and it had the highest translation score and the lowest word error rate under the same speech length. (shrink)

Existing evaluation metrics for machine translation lack crucial robustness: their correlations with human quality judgments vary considerably across languages and genres. We believe that the main reason is their inability to properly capture meaning: A good translation candidate means the same thing as the reference translation, regardless of formulation. We propose a metric that evaluates MT output based on a rich set of features motivated by textual entailment, such as lexical-semantic (in-)compatibility and argument structure overlap. We (...) compare this metric against a combination metric of four state-of-theart scores (BLEU, NIST, TER, and METEOR) in two different settings. The combination metric outperforms the individual scores, but is bested by the entailment-based metric. Combining the entailment and traditional features yields further improvements. (shrink)

The resolution of ambiguities is one of the central problems for Machine Translation. In this paper we propose a knowledge-based approach to disambiguation which uses Description Logics (dl) as representation formalism. We present the process of anaphora resolution implemented in the Machine Translation systemfast and show how thedl systemback is used to support disambiguation.The disambiguation strategy uses factors representing syntactic, semantic, and conceptual constraints with different weights to choose the most adequate antecedent candidate. We show how (...) these factors can be declaratively represented as defaults inback. Disambiguation is then achieved by determining the interpretation that yields a qualitatively minimal number of exceptions to the defaults, and can thus be formalized as exception minimization. (shrink)

Could translation be fully automated? We must first acknowledge the complexity, ambiguity, and diversity of natural languages. These aspects of natural languages, when combined with a particular dilemma known as the computational dilemma, appear to imply that the machine translator faces certain obstacles that a human translator has already managed to overcome. At the same time, science has not yet solved the problem of how human brains process natural languages and how human beings come to acquire natural language (...) understanding. We will then distinguish between the task of translation and the responsibility of the translator. Thereafter, we will conduct a survey of the methods of machine translation (viz. RBMT, SMT, NMT, foundation models or large language models). These methods will then be critically evaluated both in general and relative to Bar-Hillel’s hypothesis about the impossibility of fully automatic, high-quality machine translation (F ahqmt ). Some concluding remarks will be made about the scope, prospects, and limits of machine translation. (shrink)

Transformer is a neural machine translation model which revolutionizes machine translation. Compared with traditional statistical machine translation models and other neural machine translation models, the recently proposed transformer model radically and fundamentally changes machine translation with its self-attention and cross-attention mechanisms. These mechanisms effectively model token alignments between source and target sentences. It has been reported that the transformer model provides accurate posterior alignments. In this work, we empirically prove the (...) reverse effect, showing that prior alignments help transformer models produce better translations. Experiment results on Vietnamese-English news translation task show not only the positive effect of manually annotated alignments on transformer models but also the surprising outperformance of statistically constructed alignments reinforced with the flexibility of token-type selection over manual alignments in improving transformer models. Statistically constructed word-to-lemma alignments are used to train a word-to-word transformer model. The novel hybrid transformer model improves the baseline transformer model and transformer model trained with manual alignments by 2.53 and 0.79 BLEU, respectively. In addition to BLEU score, we make limited human judgment on translation results. Strong correlation between human and machine judgment confirms our findings. (shrink)

This book is a contribution to the research community towards thinking and reflecting on what Responsible Machine Translation really means. It was conceived as an open dialogue across disciplines, from philosophy to law, with the ultimate goal of providing a wide spectrum of topics to reflect on. It covers aspects related to the development of Machine translation systems, as well as its use in different scenarios, and the societal impact that it may have. This text appeals (...) to students and researchers in linguistics, translation, natural language processing, philosophy, and law as well as professionals working in these fields. (shrink)

Relying on large-scale parallel corpora, neural machine translation has achieved great success in certain language pairs. However, the acquisition of high-quality parallel corpus is one of the main difficulties in machine translation research. In order to solve this problem, this paper proposes unsupervised domain adaptive neural network machine translation. This method can be trained using only two unrelated monolingual corpora and obtain a good translation result. This article first measures the matching degree of (...) translation rules by adding relevant subject information to the translation rules and dynamically calculating the similarity between each translation rule and the document to be translated during the decoding process. Secondly, through the joint training of multiple training tasks, the source language can learn useful semantic and structural information from the monolingual corpus of a third language that is not parallel to the current two languages during the process of translation into the target language. Experimental results show that better results can be obtained than traditional statistical machine translation. (shrink)

On 10 February 1947, A.C. Buehler, the president of the Victor Adding Machine Company presented Norden Bombsight #4120 to the Smithsonian Institute. This sight was in service on board the Enola Gay when it dropped the first atomic bomb on Hiroshima. Through this public presentation, Buehler forever linked his company to the Norden Bombsight, the Enola Gay, and to history. Buehler’s ultimate goal, however, was the sale of adding machines, and while significant, the presentation to the Smithsonian was essentially (...) the final step in a long running advertising campaign designed to sell adding machines. During the War, Victor was the Army’s main contractor for the production of Norden Bombsights. This work is an investigation into the dysfunctional relationship that existed between Victor Adding Machine Company, the Army, the Navy Bureau of Ordinance (BuOrd). Wartime shortages demanded that pre-war arrangements between the Army and BuOrd be reconsidered and it was agreed that the Army be allowed to build its own units. Within a year of production Victor sights were scrutinized for their inaccuracies, and ultimately Victor’s contract was cancelled ending the Army’s short sojourn into bombsight production. (shrink)

Michel Galley, Pi-Chuan Chang, Daniel Cer, Jenny R. Finkel, and Christopher D. Manning Computer Science and Linguistics Departments Stanford University..

The success of deep learning in natural language processing raises intriguing questions about the nature of linguistic meaning and ways in which it can be processed by natural and artificial systems. One such question has to do with subword segmentation algorithms widely employed in language modeling, machine translation, and other tasks since 2016. These algorithms often cut words into semantically opaque pieces, such as ‘period’, ‘on’, ‘t’, and ‘ist’ in ‘period|on|t|ist’. The system then represents the resulting segments in (...) a dense vector space, which is expected to model grammatical relations among them. This representation may in turn be used to map ‘period|on|t|ist’ (English) to ‘par|od|ont|iste’ (French). Thus, instead of being modeled at the lexical level, translation is reformulated more generally as the task of learning the best bilingual mapping between the sequences of subword segments of two languages; and sometimes even between pure character sequences: ‘p|e|r|i|o|d|o|n|t|i|s|t’ → ‘p|a|r|o|d|o|n|t|i|s|t|e’. Such subword segmentations and alignments are at work in highly efficient end-to-end machine translation systems, despite their allegedly opaque nature. The computational value of such processes is unquestionable. But do they have any linguistic or philosophical plausibility? I attempt to cast light on this question by reviewing the relevant details of the subword segmentation algorithms and by relating them to important philosophical and linguistic debates, in the spirit of making artificial intelligence more transparent and explainable. (shrink)

The composition of a timeline depends on purpose, perspective, and scale – and of the very understanding of the word, the phenomenon referred to, and whether the focus is the idea or concept, an instance of an idea or a phenomenon, a process, or an event and so forth. The main function of timelines is to provide an overview over a long history, it is a kind of a mnemotechnic device or a particular kind of Knowledge Organization System (KOS).b The (...) entries in the timeline should be brief and indisputable. Therefore, timelines often identify the first occurrences rather than the most widespread or most qualified instances leaving the fuller and more complex, and possibly disputable story out. But even first occurrences are often difficult to establish. The first occurrence is most often only the first finding of an instance. Older instances may be found and competing definitions develop either within a field or in different fields. This is further complicated since the phenomena, their names, and their meanings may change over time. Former meanings may become redundant, or they must accommodate and coexist with new meanings. The time and place of the composition of the timeline are to be considered in interpreting the things listed. The following note will discuss these issues as they occur in the development of the notions of text, e-text and hypertext, and the origin of machine translation. (shrink)

This article probes the practical ethical implications of AI system design by reconsidering the important topic of bias in the datasets used to train autonomous intelligent systems. The discussion draws on recent work concerning behaviour-guiding technologies, and it adopts a cautious form of technological utopianism by assuming it is potentially beneficial for society at large if AI systems are designed to be comparatively free from the biases that characterise human behaviour. However, the argument presented here critiques the common well-intentioned requirement (...) that, in order to achieve this, all such datasets must be debiased prior to training. By focusing specifically on gender-bias in Neural Machine Translation systems, three automated strategies for the removal of bias are considered – downsampling, upsampling, and counterfactual augmentation – and it is shown that systems trained on datasets debiased using these approaches all achieve general translation performance that is much worse than a baseline system. In addition, most of them also achieve worse performance in relation to metrics that quantify the degree of gender bias in the system outputs. By contrast, it is shown that the technique of domain adaptation can be effectively deployed to debias existing NMT systems after they have been fully trained. This enables them to produce translations that are quantitatively far less biased when analysed using gender-based metrics, but which also achieve state-of-the-art general performance. It is hoped that the discussion presented here will reinvigorate ongoing debates about how and why bias can be most effectively reduced in state-of-the-art AI systems. (shrink)

This paper takes three distinct passes through the history of Machine Translation (MT) in the Soviet Union, which is typically understood as concentrating in a single boom period that lasted from roughly 1955 to 1965. In both the Soviet Union and the United States—in explicit competition with each other—there was a tremendous wave of investment in adapting computers to nonnumerical tasks that has only recently drawn the attention of historians, primarily focusing on the American example. The Soviet Union, (...) however, quickly came to assume prominence in the field both in terms of scale and diversity of approaches. At the same moment, Soviet linguists excavated a forgotten precursor, P. P. Smirnov-Troianskii, who had designed a translating machine in the early 1930s. Juxtaposing the multiple contexts in which Smirnov-Troianskii’s machine was reconceptualized and reappropriated for various ends, the article demonstrates the fundamental embodying of the algorithm in the early days of MT and also how the proliferation of narratives about Soviet MT exposes fault lines in contemporary historiography. (shrink)

Neural machine translation (NMT) is one of the success stories of deep learning and artificial intelligence. Revolutionary innovations in the computational architectures made in 2015–2017 have led to dramatic improvements in the quality of machine translation (MT) and changed the field forever. Some professional translators welcome these changes with enthusiasm, others less so. But everyone has to deal with them. Historically, the relationship between human translation and MT has been uneasy and complicated, but an increasing (...) number of players in both fields are now coming to view it as synergistic. -/- OPUS-CAT is a collection of machine translation software based on a Windows implementation of Marian NMT—a stable production-ready neural machine translation framework used in many industrial and research applications. It can be installed on your PC, fine-tuned to specialized translation domains, and integrated into CAT tools. (shrink)

… the new figure of an event-machine would no longer be even a figure. It would not resemble, it would resemble nothing, not even what we call, in a still familiar way, a monster. But it would ther...

Globalisation has triggered a proliferation of translation practises, many of which are mediated by machines. This development raises fundamental philosophical questions about language, writing, meaning, reference, and representation. This chapter builds a bridge between the ethics of machine translation and philosophy of technology. It starts by considering the activity of translation as such and argues that this is an inherently ethical activity because it involves sacrifice, establishes commonality between foreign elements, and invokes certain professional virtues. Consequently, (...) the chapter asks what machines ‘do’ to translation practises, arguing that they fundamentally transform the activity of translation into the transcription of notations. This raises the philosophical questions of logocentrism, the extent to which machines translate the ‘presence’ of lived experience, and phonocentrism, the extent to which machines transcribe the spoken word. Based on this analysis, the chapter turns to three ethical questions that pertain to machine translation. The first is about responsibility: while machines rely on retrospective responsibility, can they deal with prospective responsibility in translation? The second is about hospitality: can machines adapt to foreign worlds without having the lived experience attached to these worlds? And the third is about virtue: can the exchangeability inherent to machine translation cohere with the incommensurability of the work of translation? (shrink)

Languages with insufficient digitally available resources, such as, Indian–Indian and English–Indian language Machine Translation system developments, faces the difficulty to translate various lexical phenomena. In this paper, we present our work on a comparative study of 440 phrase-based statistical trained models for 110 language pairs across 11 Indian languages. We have developed 110 baseline statistical machine translation systems. Then, we have augmented the training corpus with Indowordnet synset word entries of lexical database and further trained 110 (...) models on top of the baseline system. We have done a detailed performance comparison using various evaluation metrics such as BLEU score, METEOR, and TER. We observed significant improvement in evaluations of translation quality across all the 440 models after using the Indowordnet. These experiments give a detailed insight in two ways: usage of lexical database with synset mapping for resource poor languages and efficient usage of Indowordnet synset mapping. Moreover, synset mapped lexical entries helped the SMT system to handle the ambiguity to a great extent during the translation. (shrink)

During my recent work on machine translation, I have come across many problems of a linguistic nature that should be of general methodological interest. Some of these problems have never been treated extensively before. Others that have been discussed previously appear now in a different and rather interesting light.

Topic model is a well proven tool to investigate the semantic content of textual corpora. Yet corpora sometimes include texts in several languages, making it impossible to apply language-specific computational approaches over their entire content. This is the problem we encountered when setting to analyze a philosophy of science corpus spanning over eight decades and including original articles in Dutch, German and French, on top of a large majority of articles in English. To circumvent this multilingual problem, we use (...) class='Hi'>machine-translation tools to bulk translate non-English documents into English. Though largely imperfect, especially syntactically, these translations nevertheless provide correctly translated terms and preserve the semantic proximity of documents with respect to one another. To assess the quality of this translation step, we develop a “semantic topology preservation test” that relies on estimating the extent to which document-to-document distances have been preserved during translation. We then conduct an LDA topic-model analysis over the entire corpus of translated and English original texts, and compare it to a topic-model done over the English original texts only. We thereby identify the specific contribution of the translated texts. These studies reveal a more complete picture of main topics that can found in the philosophy of science literature, especially during the early days of the discipline when numerous articles were published in languages other than English. (shrink)

Linking constructions involving dሇ (DE) are ubiquitous in Chinese, and can be translated into English in many different ways. This is a major source of machine translation error, even when syntaxsensitive translation models are used. This paper explores how getting more information about the syntactic, semantic, and discourse context of uses of dሇ (DE) can facilitate producing an appropriate English translation strategy. We describe a finergrained classification of dሇ (DE) constructions in Chinese NPs, construct a corpus (...) of annotated examples, and then train a log-linear classifier, which contains linguistically inspired features. We use the DE classifier to preprocess MT data by explicitly labeling dሇ (DE) constructions, as well as reordering phrases, and show that our approach provides significant BLEU point gains on MT02 (+1.24), MT03 (+0.88) and MT05 (+1.49) on a phrasedbased system. The improvement persists when a hierarchical reordering model is applied. (shrink)

Pi-Chuan Chang, Michel Galley, and Christopher D. Manning Computer Science Department, Stanford University Stanford, CA 94305 pichuan,galley,[email protected]..

The advantages and disadvantages of machine translation have been the subject of increasing debate among human translators lately because of the growing strides made in the last year by the newest major entrant in the field, Google Translate. The progress and potential of machine translation has been debated much through its history. But this debate actually began with the birth of machine translation itself. Behind this simple procedure lies a complex cognitive operation. To decode (...) the meaning of the source text in its entirety, the translator must interpret and analyze all the features of the text. The success of encoding of war rivals texts during WWII pushed the language experts to believe that translation of foreign languages could be realized based on these encoding principles. The cold war coincided with the invention of computers, and “cracking Russian” was one of the first tasks these machines were set. The competition towards establishing more business with different parts of the world incited advanced countries in technology to look for easy and quick ways for communication. Nevertheless, machine translation has proved helpful in more urgent situations as well. But, in my opinion, although computer scientists have toiled for decades to produce machine translation comparable to that rendered by humans, they have yet to succeed. (shrink)

At the level of English resource vocabulary, due to the lack of vocabulary alignment structure, the translation of neural machine translation has the problem of unfaithfulness. This paper proposes a framework that integrates vocabulary alignment structure for neural machine translation at the vocabulary level. Under the proposed framework, the neural machine translation decoder receives external vocabulary alignment information during each step of the decoding process to further alleviate the problem of missing vocabulary alignment (...) structure. Specifically, this article uses the word alignment structure of statistical machine translation as the external vocabulary alignment information and introduces it into the decoding step of neural machine translation. The model is mainly based on neural machine translation, and the statistical machine translation vocabulary alignment structure is integrated on the basis of neural networks and continuous expression of words. In the model decoding stage, the statistical machine translation system provides appropriate vocabulary alignment information based on the decoding information of the neural machine translation and recommends vocabulary based on the vocabulary alignment information to guide the neural machine translation decoder to more accurately estimate its vocabulary in the target language. From the aspects of data processing methods and machine translation technology, experiments are carried out to compare the data processing methods based on language model and sentence similarity and the effectiveness of machine translation models based on fusion principles. Comparative experiment results show that the data processing method based on language model and sentence similarity effectively guarantees data quality and indirectly improves the algorithm performance of machine translation model; the translation effect of neural machine translation model integrated with statistical machine translation vocabulary alignment structure is compared with other models. (shrink)