New Dilemmas in Corpus-Based Translation Studies

This chapter discusses the implication of corpus-based translation studies as a new research methodology in translation studies, which expands the scope of translation studies. The chapter then analyzes factors to be investigated in corpus-based translation studies following an analysis of the problems that arise. Despite all the limitations, corpus-based translation studies promise to be a worthy field for research in translation studies.

Several methodologies have recently been proposed to enhance the performance of low-resource Neural Machine Translation (NMT). However, these techniques have yet to be explored thoroughly in the low-resource Thai and Myanmar languages. Therefore, we first applied augmentation techniques such as SwitchOut and Ciphertext Based Data Augmentation (CipherDAug) to improve NMT performance in these languages. Second, we enhanced the NMT performance by fine-tuning the pre-trained Multilingual Denoising BART model (mBART), where BART denotes Bidirectional and Auto-Regressive Transformer. We implemented three NMT systems: namely, Transformer+SwitchOut, Multi-Source Transformer+CipherDAug, and fine-tuned mBART in the bidirectional translations of Thai-English-Myanmar language pairs from the ASEAN-MT corpus. Experimental results showed that Multi-Source Transformer+CipherDAug significantly improved Bilingual Evaluation Understudy (BLEU), Character n-gram F-score (ChrF) , and Translation Error Rate (TER) scores over the first baseline Transformer and second baseline Edit-Based Transformer. The model achieved notable BLEU scores: 37.9 (English-to-Thai), 42.7 (Thai-to-English), 28.9 (English-to-Myanmar), 31.2 (Myanmar-to-English), 25.3 (Thai-to-Myanmar), and 25.5 (Myanmar-to-Thai). The fine-tuned mBART model also considerably outperformed the two baselines, except for the Myanmar-to-English pair. SwitchOut improved over the second baseline in all pairs and performed similarly to the first baseline in most cases. Last, we performed detailed analyses verifying that the CipherDAug and mBART models potentially facilitate improving low-resource NMT performance in Thai and Myanmar languages.

The competitive performance of neural machine translation (NMT) critically relies on large amounts of training data. However, acquiring high-quality translation pairs requires expert knowledge and is costly. Therefore, how to best utilize a given dataset of samples with diverse quality and characteristics becomes an important yet understudied question in NMT. Curriculum learning methods have been introduced to NMT to optimize a model's performance by prescribing the data input order, based on heuristics such as the assessment of noise and difficulty levels. However, existing methods require training from scratch, while in practice most NMT models are pre-trained on big data already. Moreover, as heuristics, they do not generalize well. In this paper, we aim to learn a curriculum for improving a pre-trained NMT model by re-selecting influential data samples from the original training set and formulate this task as a reinforcement learning problem. Specifically, we propose a data selection framework based on Deterministic Actor-Critic, in which a critic network predicts the expected change of model performance due to a certain sample, while an actor network learns to select the best sample out of a random batch of samples presented to it. Experiments on several translation datasets show that our method can further improve the performance of NMT when original batch training reaches its ceiling, without using additional new training data, and significantly outperforms several strong baseline methods.

In this paper, we improve the performance of neural machine translation (NMT) with shallow syntax (e.g., POS tag) of target language, which has better accuracy and latency than deep syntax such as dependency parsing. We present three NMT decoding models (independent decoder, gates shared decoder and fully shared decoder) to jointly predict target word and POS tag sequences. Experiments on Chinese-English and German-English translation tasks show that the fully shared decoder can acquire the best performance, which increases the BLEU score by 1.4 and 2.25 points respectively compared with the attention-based NMT model.

Due to the lack of structured language resources, low-resource language machine translation often faces difficulties in cross-language semantic paraphrasing. In order to solve the problem of low-resource machine translation from Indonesian to Chinese, a cognate-parallel-corpus-based expanding method of language resources is proposed, and an improved neural machine translation model is trained by the Malay-corpus-enhanced corpus. The improved model can achieve a comparable result as that of Google in the experiment of Indonesian-Chinese machine translation. The experimental results also show that the morphological similarity and semantic equivalence between the languages are very effective computational features to improve the performance of neural machine translation for low-resource languages.

Pre-training (PT) and back-translation (BT) are two simple and powerful methods to utilize monolingual data for improving the model performance of neural machine translation (NMT). This paper takes the first step to investigate the complementarity between PT and BT. We introduce two probing tasks for PT and BT respectively and find that PT mainly contributes to the encoder module while BT brings more benefits to the decoder. Experimental results show that PT and BT are nicely complementary to each other, establishing state-of-the-art performances on the WMT16 English-Romanian and English-Russian benchmarks. Through extensive analyses on sentence originality and word frequency, we also demonstrate that combining Tagged BT with PT is more helpful to their complementarity, leading to better translation quality. Source code is freely available at https://github.com/SunbowLiu/PTvsBT.

Apart from its contribution to the analysis of translated discourse as such, corpus-based translation studies has often involved the comparison of translated corpora and comparable originals, in an attempt to isolate the features that typify translations, whether globally or in a more restricted set. The study reported here applied a similar methodology to the analysis of interpreted discourse, comparing it not to non-interpreted (spontaneous, original) spoken discourse but to its written (translated) counterpart. A computerized analysis of the interpreted outputs of six professional translator-interpreters rendering the same text from their second to their first language in both modalities revealed a set of marked differences between them in terms of richness (type-token ratio), and of a range of lexico-grammatical features. Despite its drawbacks in terms of ecological validity, the methodology used in this study is seen as a tool for extrapolating a set of stylistic and pragmatic features of interpreted – as opposed to translated – outputs, and may constitute an extension of the range of the paradigms available to corpus-based translation studies. A statistical analysis of the morphological data generated pointed to salient differences between the two corpora, and it is these differences that are at the core of the present study. The methodological implications and possible extensions are also discussed below.

Neural Machine Translation (NMT) has recently achieved the state-of-the-art in many machine translation tasks, but one of the challenges that NMT faces is the lack of parallel corpora, especially for low-resource language pairs. And the result is that the performance of NMT is much less effective for low-resource languages. To address this specific problem, in this paper, we describe a novel NMT model that is based on encoder-decoder architecture and relies on character-level inputs. Our proposed model employs Convolutional Neural Networks (CNN) and highway networks over character inputs, whose outputs are given to an encoder-decoder neural machine translation network. Besides, we also present two other approaches to improve the performance of the low-resource NMT system much further. First, we use language modeling implemented by denoising autoencoding to pre-train and initialize the full model. Second, we share the weights of the front few layers of two encoders between two languages to strengthen the encoding ability of the model. We demonstrate our model on two low-resource language pairs. On the IWSLT2015 English-Vietnamese translation task, our proposed model obtains improvements up to 2.5 BLEU points compared to the baseline. We also outperform the baseline approach more than 3 BLEU points on the CWMT2018 Chinese-Mongolian translation task.

Recently, neural machine translation (NMT) has made a remarkable achievement. However, performance of NMT is highly influenced by the size of training parallel text. The required amount of parallel text is not available for low-resource languages. The issue of low-resource NMT exacerbated if at least one of paired languages is morphologically rich. Inspired by the capabilities of pre-trained word embedding models, we propose simple yet effective word embedding based parallel text augmentation technique. The proposed technique is vital to enlarge size of limited parallel text for low-resource NMT. To this end, we adopt publicly available pre-trained word embedding models that favor morphologically rich languages. Accordingly, we replace each word in the original text with the n most similar words in the vector space (in this study n=3). For experimental analysis, we simulate low-resource languages by using publicly available parallel texts for German↔English, Turkish↔English and Finnish↔English language pairs. We use the original parallel texts as a seed for generating synthetic parallel texts and for training character-level NMT that deemed as baseline models. The obtained results show that our augmentation techniques improve performance of baseline and dummy source sentence based NMT models by up to 5.9 CHRF scores.

Morphologically rich and complex languages such as Arabic, pose a major challenge to neural machine translation (NMT) due to the large number of rare words and the inability of NMT to translate them. Unknown word (UNK) symbols are used to represent out-of-vocabulary words because NMT typically operates with a fixed vocabulary size. These rare words can be effectively encoded as sequences of subword units by using algorithms, such as byte pair encoding (BPE), to tackle the UNK problem. However, for languages with highly inflected and morphological variations, such as Arabic, the aforementioned method has its own limitations that make it not effective enough for translation quality. To alleviate the UNK problem and address the inconvenient behavior of BPE when translating the Arabic language, we propose to utilize a romanization system that converts Arabic scripts to subword units. We investigate the effect of our approach on NMT performance under various segmentation scenarios and compare the results with systems trained on original Arabic form. In addition, we integrate Romanized Arabic as an input factor for Arabic-sourced NMT compared with well-known factors, namely, lemma, part-of-speech tags, and morph features. Extensive experiments on Arabic-Chinese translation demonstrate that the proposed approaches can effectively tackle the UNK problem and significantly improve the translation quality for Arabic-sourced translation. Additional experiments in this study focus on developing the NMT system on Chinese-Arabic translation. Before implementing our experiments, we first propose standard criteria for the data filtering of a parallel corpus, which helps in filtering out its noise.

Corpus-based translation studies (CTS) is a very new technique in translation studies. In Bible translation, researchers have been collating corpora of source texts since the eighteenth century. In recent times, corpora of translated Bible texts have also been generated by a number of researchers. Thus the notion of working with corpora rather than with individual texts is not new to Bible translation. However, until very recently these corpora were examined manually, so that the introduction of computer-aided tools for corpus analysis represents a significant advancement for the Bible translator or researcher. By its very nature, the Bible lends itself remarkably well to (computerised) corpus projects. It is by definition a corpus of 66 books, and the pre-existing divisions into books, chapters and verses gives it an unusual advantage in annotation over literary works in the context of a parallel corpus. This article looks at the ways in which the Bible and its translations lend themselves to the formation of a specific corpus, the types of corpora and texts that could be considered as well as the challenges involved in producing these parallel corpora. The article then goes on to examine the potential applications of such a corpus for both the production of new Bible translations and the evaluation of existing ones. Applications to other nonreligious linguistic problems are also explored. In assessing these challenges, it becomes evident that CTS indeed redefines the nature of translation research as well as the role of the researcher. Finally, the status of existing projects to produce a parallel corpus based on Biblical texts will be reported.

As the literature on COVID-19-themed translation continues to expand, the area of corpus-based cultural translation in this context still needs to be explored. This study investigates the types of information sharing and new databases that have emerged in response to the COVID-19 pandemic and assesses the feasibility of human-machine interaction in the selection of translation strategies. Our study employs a systematic approach to analyze and compare the linguistic characteristics of epidemic-related English texts with those of general English texts. Using a keyword extraction method, we identify unique vocabulary features in each text type and then focus on the most salient features specific to the epidemic-related context. Drawing from COVID-19 and SARS-CoV-2 related research, the study examines the importance of texts in complex social situations, their influential role, and the potential for human-machine collaboration in translation strategy selection. The findings suggest combining functional equivalence theory and corpus analysis methods can effectively enhance cross-cultural communication for epidemic-related themes. This approach leads to improved translation quality through increased cohesion and coherence. The research delves deeper into the translation process by building upon the insights gained from the corpus study and adopting a communicative perspective. It goes beyond simply identifying equivalent terms and tackles the transfer of discursive phenomena, such as collocations, terminological variations, and semantic prosody from the source to the target language. The findings of this study confirm that corpus-based translation studies' contribution to ESP implies moving towards a communicative and socially oriented approach to specialized language.

The pervasiveness of corpus-based research in the broad interdisciplinary field of translation studies is well attested. This editorial briefly reports on some of the most significant academic initiatives undertaken in corpus-based translation studies in recent years. It introduces each of the eight papers selected for this special issue of Research in Corpus Linguistics (RiCL). In doing so, the editorial will highlight their distinctive contribution to the interdisciplinarity of translation and interpreting studies.

This chapter discusses the role played by corpus linguistics in contemporary translation studies in past decades. We will begin with an introduction to the fundamental reformative influence of corpus linguistics upon the methodologies of the contemporary linguistic research, which is then followed by a more detailed description of the most frequently used types of corpora, particularly multilingual corpora, in contrastive linguistic research and translation studies. We will also introduce a growing paradigm of research by exploring the latest developments in corpus-based translation studies.

The advent of Generative AI (Gen-AI) has precipitated an epistemological crisis in strategic management, challenging the foundational assumptions of enterprise digital transformation. Traditional frameworks, operating under a “Paradigm of Rational Mastery”, view technology as a predictable, manageable tool. This paper argues that the unique technical characteristics of Gen-AI—specifically its “Emergent Abilities” and “Black-Box Nature”—have given rise to a novel strategic condition termed “Foundational Uncertainty”, rendering these established models obsolete. This study critiques the prevalent “Two-Layer Model” of digital transformation, which is predicated on predictable technological advancement and planned “enablement”. In its place, this paper develops and proposes a “Dynamic-Adaptive Model” designed for the Gen-AI era. This new model reconceptualizes the firm-technology relationship as a co-evolutionary process. It posits that firms must shift from planned implementation to continuous “Adaptive Sensemaking”, an action-driven process of experimentation and learning to navigate the technology’s unpredictability. At the enterprise level, strategy must become an “Evolutionary Process” that emerges from this learning. The model’s central mechanism is a cyclical interplay of “Contention & Adaptation”—a generative tension between top-down managerial control (exploitation) and bottom-up, technology-driven learning (exploration) that propels organizational evolution. Consequently, the source of sustained competitive advantage shifts from possessing static, inimitable (VRIN) resources to cultivating a higher-order “Dynamic Adaptive Capability”—the organizational meta-capability for reconfiguring its own learning processes to thrive amidst unknowable change. The paper concludes by outlining the model’s significant theoretical contributions to the literature on digital transformation, Dynamic Capabilities, and the Resource-Based View, providing a new framework for navigating the Gen-AI frontier.