This special issue of Across Languages and Cultures stems from the seventh edition of the Using Corpora in Contrastive and Translation Studies (UCCTS) conference, held this time in Poznań in 2023. Since its launch in 2008 by Richard Xiao, this biennial international event offers a global forum for the exploration of theoretical and practical issues related to the development and application of corpora in contrastive and translation/interpreting studies. It is also well known as one of the few events in translation studies that focuses on methodology: in this case, corpus linguistics, integrated with a range of natural language processing tools to investigate the underlying linguistic phenomena in various contexts of language use, particularly language mediation.
This marks the second time this journal has published a special issue following the UCCTS event. The first was nearly a decade ago, featuring papers from the 2013 conference in Ghent, guest-edited by Bart Defrancq, Bernard De Clerck, and Gert De Sutter (2015).
Back then, much of the focus in the Corpus-based Translation and Interpreting Studies (CTIS) was on testing translation universals like simplification and explicitation. This emphasis was reflected in the fact that the opening section of the introductory paper of that issue was titled The road often taken: testing universals in written corpora (Defrancq, De Clerck, & De Sutter, 2015, p. 157). Several years later, Granger and Lefer (2022, p. 36), in their review of the Corpus-based Translation Studies (CBTS) papers published in the leading translation journals in 2012–2019, observe that while translation features like explicitation still remain a rather popular subject of investigation, they emerged as keywords in only 29% of the examined abstracts. This leaves over 70% to other topics, which shows that the field has progressed. Constantly extending its scope, researchers investigate increasingly more intricacies of the multifaceted nature of linguistic data in new forms of language mediation.
The fact that the concept of translation keeps evolving has been pointed out by scholars such as Gambier and Kasperẹ (2021), who report that emerging translation practices like localisation, amateur translation, and media translation are reshaping the field. With technological advancements affecting translation workflows, they suggest that novel translator roles are emerging, blurring the boundaries of traditional translation concepts. This ambiguity warrants fresh definitions of such interrelated basic notions as ‘text’. Needless to say, these changes also drive the advancement of research methods that could adequately describe them.
Even though advances in natural language processing tools have enabled corpus studies to probe linguistic nuances in greater detail than ever before, Granger and Lefer (2022, p.34) observe that “the majority of the corpus-based translation and interpreting studies in [their] survey do not exploit the full potential offered by the electronic nature of the corpus”.
The need for methodological advancement could be seen at a special roundtable discussion at the conference in Poznań, convened by Silvia Bernardini, Stella Neumann, and Gert De Sutter as part of the INTERACT research network (funded by the Research Foundation Flanders, grant number W002220N). The roundtable focused on innovative corpus research designs for 21st-century contrastive and translation studies. As the panel's convenors rightly pointed out, these designs need to respond to the changing character of translation with less “clear-cut boundaries between what counts as an ‘original’ and a ‘translation’, massively distributed collaborative processes, and human translation ‘augmented’ by machine translation in various ways” (Bernardini, Neumann, & De Sutter, 2023).
The panellists thus explored a range of issues to address this new complexity, including novel corpus designs for translation and contrastive studies, moving beyond the conventional comparable/parallel dichotomy, addressing the challenges of multi-parallel corpora, balancing sampling bias against coverage in translation-related texts, ensuring ecological validity versus elicitation, achieving cross-linguistic comparability of data sets, leveraging rich metadata, and using corpora to study behaviour and cognition. Several of these topics were addressed in presentations at the conference.
Overview of the papers in this issue
The papers selected for this special issue were curated to showcase some of these central themes. As evidenced by this special issue, CTIS researchers persist in exploring recurrent linguistic patterns, yet the approaches, methods, and tools have undergone significant evolution. There is growing interest in building and analysing corpora for non-canonical forms of translation, such as news translation, indirect translation, machine translation, and sign language interpreting, which are addressed in this issue. At the same time, studies on more traditional forms of translation, like literary texts, or intermodal comparisons between translation and interpreting, employ innovative and increasingly automatized data processing methods and multifactorial analyses.
Silvia Bernardini, Adriano Ferraresi, Federico Garcea and Natalia Rodriguez Blanco take on the complex nature of news translation, where conventional corpus analysis encounters challenges caused by radical rewriting and close translation. They test two novel methods to isolate translated segments from the news items compiled in a trilingual Agence France-Press corpus. One leverages machine translation with semantic similarity scores, while the other employs multilingual sentence embeddings. They construct a distinctive “comparallel” corpus that combines parallel and comparable corpus architectures. This pioneering methodology opens original avenues for news translation research, paving the way for further exploration of novel translation forms.
Ilmari Ivaska and Laura Ivaska scrutinise the linguistic cues distinguishing source languages in direct and indirect translations into Finnish. Employing Halverson's (2017) gravitational pull model as a theoretical background, they utilise supervised and unsupervised machine learning techniques to identify pivotal linguistic traits for this distinction. The findings reveal that features such as sentence length, sentence-initial adverbs, and sentence-final specification are key in demarcating the source languages. Moreover, the prominent characteristics of the original source language overshadow those of the intermediary languages in indirect translations or the entrenched similarities between language pairs.
Daniel Henkel studies passive constructions in English and French, analysing a bidirectional corpus of original and translated texts produced by human translators and machine translation. While passive constructions were more frequent in original English compared to French, translated texts showed fewer disparities, except for French translations by DeepL. Machine translation exhibited a stronger correlation with source texts, indicating a higher imitation level than human translation. Manual analysis confirmed these findings, revealing more consistent translation patterns in machine translation. Overall, this study elucidates the divergent translation patterns between human translators and machine systems, revealing nuanced structural preferences.
Silvia Gabarro-Lopez, Laurence Meurant and Nicolas Hanquet present an exhaustive description of the French Belgian Sign Language (LSFB) corpus, a comprehensive multilingual resource encompassing French and LSFB parallel and comparable corpora. The paper outlines the corpus structure and its compilation, stressing at the same time its relevance for contrastive inquiry of sign language interpreting studies. It also reports on a novel study comparing reformulation structures in semi-spontaneous LSFB dialogues and co-interpreted LSFB data, illustrating the potential for further comparison between these corpora to uncover previously unexplored nuances of co-interpretation. The study illuminates distinctive aspects of sign language interpreting and shows novel research trajectories.
Christina Pollkläsener, Frances Yung and Ekaterina Lapshinova-Koltunski delve into discourse markers and connectives, scrutinising written and spoken texts originally produced in English, as well as English translations and interpretations from German. Employing automated annotation, they examine the distributions of these features across the diverse language production modes. The study discerns patterns of explicitation and implicitation, linking them to the cognitive complexity of discourse relations. This offers valuable insights into the differential treatment of connectives in interpreting and translation depending on discourse relations.
Lorenzo Mastropierro and Łukasz Grabowski report on a corpus-based study identifying factors influencing repetition in the translation of reporting verbs from English to Italian and Polish in literary translation. They employ multiple negative binomial regression models to analyse the effects of five predictor variables on the number of different target language verb types into which a reporting verb in the source text is translated. The results indicate that factors such as verb category, frequency of source-text reporting verb, and number of translation equivalents significantly contribute to explaining variation in the response variable. The study proposes a novel approach to analysing repetition in mediated texts and provides a detailed understanding of the nature of repetition of reporting verbs in literary translation.
This selection of studies aims to showcase current trends in CTIS while offering insights into the new or rarely explored aspects of language mediation. It also utilizes state-of-the-art data extraction and analysis methods to shed new light on more traditional forms of language mediation.
A look ahead
What are the possible future trajectories? As with any research field dependent on technology, Corpus-based Translation and Interpreting Studies is dynamic and continually explores novel methods and research avenues. Like in Corpus Linguistics (CL), there is a gradual shift towards more advanced statistical analyses in CTIS, although this transition appears to be more rapid in the former. Larsson, Egbert, and Biber (2022, p. 150) observe that in CL, this shift has come “at the expense of linguistic description and analysis”; hence, it is crucial for CTIS scholars to maintain a reasonable balance between these aspects.
Data collection, processing, and analysis in any field reliant on large linguistic data collections will likely become increasingly automated with artificial intelligence (AI). AI will enable researchers, even those with minimal natural language processing background, to collect and examine larger datasets, making their outcomes more generalisable. Particularly in this context, fostering a culture of transparency and sharing data and analysis files would help the discipline harness AI's full potential, while mitigating its risks, such as system opacity, biases, and variability of outcomes.
The relatively recent trend towards augmenting corpora with very rich metadata, which has clear benefits even in the case of more traditional corpora (Reynaert, Macken, Tezcan, & De Sutter, 2021), may help maintain the transparency necessary, particularly in the context of AI and facilitate data exchange and collaboration.
Fortunately, the collaborative efforts within the CTIS community and the regular exchange of ideas at events like UCCTS promise continued progress, ensuring that this research field remains at the forefront of linguistic inquiry.
Acknowledgements
The guest editor would like to thank the authors whose contributions are featured in this special issue and the anonymous reviewers for their valuable feedback on the submitted manuscripts. Appreciation is also due to the journal editors for making this special issue possible and their steady support and careful oversight of the publication process. Finally, appreciation goes to the UCCTS community members for their thought-provoking discussions during the conference and the INTERACT research network for organising an exceptionally engaging roundtable session on innovative corpus-based research designs.
References
Bernardini, S., Neumann, S., & De Sutter, G. (2023, July 10–12). Towards enriched corpus-based research designs for 21st century contrastive and translation studies [Roundtable Discussion]. Using Corpora in Contrastive and Translation Studies. https://wa.amu.edu.pl/uccts2023/roundtable_discussion.html.
Defrancq, B., De Clerck, B., & De Sutter, G. (2015). Corpus-based translation studies: Across genres, methods and disciplines. Across Languages and Cultures, 16(2), 157–162.
Gambier, Y., & Kasperẹ, R. (2021). Changing translation practices and moving boundaries in translation studies. Babel, 67(1), 36–53.
Granger, S., & Lefer, M. A. (2022). Corpus-based translation and interpreting studies. In M. A. Lefer, & S. Granger (Eds.), Extending the scope of corpus-based translation studies (pp. 13–41). Bloomsbury: Bloomsbury Advances in Translation series.
Halverson, S. (2017). Gravitational pull in translation. Testing a revised model. In G. De Sutter, M.-A. Lefer, & I. Delaere (Eds.), Empirical translation studies: New methodological and theoretical traditions (pp. 9–46). De Gruyter.
Larsson, T., Egbert, J., & Biber, D. (2022). On the status of statistical reporting versus linguistic description in corpus linguistics: A ten-year perspective. Corpora, 17(1), 137–157.
Reynaert, R., Macken, L., Tezcan, A., & De Sutter, G. (2021). Building a new-generation corpus for empirical translation studies: The Dutch Parallel Corpus 2.0. In New perspectives on corpus translation studies (pp. 75–100). Springer.
