Entretien
SPOTLIGHT ON… FERNANDO PRIETO RAMOS
Fernando Prieto Ramos is full professor and director of the Transius Centre at the University of Geneva's Faculty of Translation and Interpreting, where he held the deanship from 2014 to 2018 and is currently vice-dean. His work focuses on interdisciplinary approaches and quality assurance in legal and institutional translation in particular. He has published widely, and has received several academic awards, including an International Geneva Award from the Swiss Network for International Studies and a Consolidator Grant for his project on “Legal Translation in International Institutional Settings” (LETRINT). He is a former member of the Centre for Translation and Textual Studies at Dublin City University, and has translated for various institutions since 1997, including five years of in-house service at the World Trade Organization (dispute settlement team). In this interview, he tells us more about the LETRINT project.
Can you briefly tell us about the LETRINT project, its origins and its aims?
LETRINT (“Legal Translation in International Institutional Settings: Scope, Strategies and Quality Markers”) is a Consolidator Grant project, the first of its kind in Translation Studies in Europe. It applies an evidence-based approach to define the scope of the field, describe its main features and examine aspects of translation quality from a holistic perspective that encompasses process, competence and product.
This project was partly motivated by my experience as a legal translation practitioner both in national and international settings. I had come across quality issues that ought to be addressed in specialized training, particularly when dealing with legal terminology and the typical problems of incongruity between legal systems. I had also dealt with a wide variety of legal genres that were not covered by works on legal translation or institutional translation. When I came up with the project in the mid-2010s, this field was attracting increasing attention, but was still under-researched and predominantly focused on translation for the EU institutions and its legislative genres in particular. Although the international organizations were traditionally regarded as centres of reference in the professionalization of translation – and in spite of the improved accessibility of institutional text repositories –, the broader scope and specific features of translation in these institutions had yet to be examined empirically.
As a researcher, I had worked on a multi-dimensional approach to decision-making and a competence model in legal translation, so exploring institutional practices in this regard was a natural continuation of my work and in line with the strategic priorities of our Faculty and Centre for Legal and Institutional Translation Studies (Transius).
What challenges did you face during this project and what solutions did you come up with?
The project had to overcome multiple challenges. The most significant one was its ambitious scope. In practice, it was like three large projects in one, which meant managing several workstreams simultaneously. I was Faculty dean when the project was launched, which represented another major challenge for time management and eventually motivated my not running for a second term. In the end, given the progress and level of ambition of the project, the initial 5-year duration was extended for an additional 21 months.
For the first phase, corpus building was a huge undertaking. I had promised a full mapping of text production in three representative institutional settings (the main EU institutions, the UN and the WTO) and three common languages (English, French and Spanish). Several technical solutions for bulk downloading and tagging were necessary to compile LETRINT’s first set of corpora (LINST), made up of 1.71 billion words. All translated texts from LINST then had to be classified according to several key functions (law- and policy-making, implementation monitoring, adjudication and administrative functions), for comparative purposes, to build the LETRINT 0 parallel corpus, which was a very complex and time-consuming process.
Another major difficulty was figuring out how to preserve the representativeness of our datasets while downsizing the corpus to make it more manageable for alignment and reliable human annotation. This was achieved through stratified sampling techniques that took into consideration quantitative and qualitative criteria. This kind of approach had no precedent in the field, so it was ground-breaking (and risky), but fortunately proved to be successful, yielding the LETRINT 1 and LETRINT 1+ corpora of 7.72m and 0.75m words, respectively. More information on the design and features of the LETRINT corpora can be found on the project website, including an infographic and overview of the various datasets.
Simultaneously, we had to collaborate with an IT developer to adapt a concordancer to exploit the LETRINT 1 and LETRINT 1+ trilingual corpora, since there was no tool on the market that fully met the project’s needs. The resulting query tool, “LETRINT-Q”, is also available online.
When implementing the project, another challenge involved coordinating the research team to apply the criteria for corpus analysis in a uniform manner, for example, for classifying texts according to legal functions and genres, assessing translation adequacy levels, or annotating text segments. In order to ensure the validity of data and reduce the risk of subjective distortions, team members were carefully selected for their profiles, and I built a number of safeguards into the workflows, including double validations, iterative refinements of categories and consensus-building mechanisms to resolve problematic or borderline classifications. This process added real value that could not be matched by automated annotations. I was able to put together an excellent team to contribute to these tasks and I am very grateful to all of them for their good work.
I should also mention the challenge of triangulating data that was obtained from many source types and institutions. It not only included the text corpora, but also the data from structured interviews (more than 30 with 45 service managers or quality advisers), surveys and institutional materials (statistics, policy documents, terminological resources, vacancy notices, guidelines, etc.). Processing these data was fascinating but also complex, given the multiple quantitative and qualitative aspects that had to be analyzed from a comparative and sometimes diachronic perspective. While the analysis of the LETRINT corpora and interviews focused on three settings, some of the surveys and studies (e.g., on translators’ profiles, job descriptors or the use of tools) involved up to 24 international organizations.
What were the results of the project and what contribution have they made to the field of institutional translation?
Given their unique features, the LETRINT corpora in and of themselves are a major contribution that can be of value for other corpus-based studies. They are freely accessible online. The full textual and contextual mapping led to an empirical description of the vast scope of institutional legal translation, in terms of multilingual text production, according to legal function and genre. The mapping also enabled us to refine a matrix for text categorization in the field and capture the commonalities and variations between the contexts under analysis. The triangulation of the textual mapping with survey results confirmed that legal translation is most prominent in the Court of Justice of the EU (CJEU), the international courts and the EU law-making institutions, compared to the UN as a whole and other intergovernmental organizations, where our mapping revealed larger translation volumes in monitoring procedures. Yet, the legal specialization in translation is the most prevalent across the board.
Our findings also point to a clear relationship between the number of official languages and the variations of multilingualism, as implemented in each legal setting. The data obtained provide evidence of the underlying legal, textual and linguistic hierarchies that prevail in translation directionality, strategies and quality assurance, and in multilingual text interpretation. Aside from the CJEU, where French remains the main working language, English is the prevailing lingua franca in all the settings, and the key law-making and adjudicative text genres, such as regulations and judgments, are amongst the most revised and least frequently outsourced, in line with the new trend toward more structured approaches to translation quality assurance, particularly in the EU institutions. The role of translation service heads in implementing quality assurance was also examined based on information gathered from the interviews and job descriptors.
As a distinctive feature of legal texts and an indicator of both competence and quality in legal translation, terminology has been used as a key marker to connect process, competence and product. As part of the analysis of discourse features and, in particular, the density of several categories of terminology and phraseology, LETRINT has revealed a high degree of discursive and thematic hybridity in key genres of EU and international law. The most significant differences lie in the terminology of non-legal domains, in accordance with the main policy areas of each institution. For example, more frequent economic terminology can be found in WTO texts.
The findings of our corpus annotation also provide the first empirical evidence of its kind on the difficulty levels associated with terminology when translating international legal discourses. This also contributes to the broader debate of what can be considered a “problem” in translation, depending on the translator profile and the cognitive effort required for decision-making. According to the difficulty scores obtained, the most difficult category for translation purposes is national legal terminology, especially when it comes to terms designating singular entities and concepts.
These terms are often associated with issues of legal asymmetry that call for solid legal translation competence. The correlation between legal singularity and translation difficulty also aligns with the results of our analysis of the adequacy and consistency of translations of illustrative legal terminology, and the gaps detected in legal term records of institutional terminological databases. A more recent study of neural machine translations of the same terminology (to be published next year) points to similar difficulties. Terminological issues are also among the translation errors scrutinized in our analysis of corrigenda issued by international organizations, a highly relevant source for learning about institutional translation quality control and formal correction processes.
As for the connections between legal translation competence, decision-making and product quality, we have found, for example, significant profile variations in the patterns of use and perceptions of resources for legal translation. From a broader and more diachronic perspective, a comparative analysis of vacancy notices for translator and reviser positions has explored competence requirements and recruiting approaches across institutions, including domain specialization and the gradual integration of technological skills in job descriptors. Most recently, we have conducted a survey on the use of tools and the impact of neural machine translation on translation processes, skills and quality, as perceived by institutional translation professionals. These results, along with the findings of our tests on the impact of domain specialization on quality control performance, will be publicly available in the coming months. For this last phase, a quality evaluation framework has been developed following a comprehensive review of professional, academic and institutional approaches and standards. You can find more details about the project outputs on the LETRINT website.
Overall, LETRINT’s results contribute to a more advanced and nuanced understanding of the aspects mentioned earlier, and can support specialized training and institutional improvements, including resource and professional development, workflow management, recruitment policies and other quality enhancement measures. Our early findings in connection with term banks, for example, have already contributed to new initiatives to revise legal term records in UNTERM.
What other projects are you involved in?
I have been involved in the European Strategic Partnership project EFFORT (“European Framework for Translation”), as a member of the management committee and coordinator of the working group on legal translation. I have also been a member of the management committee of the LITHME COST project (“Language in the Human-Machine Era”) since its conception, and currently co-chair its working group on “Law and language”. In recent years, I have also participated in new research collaborations on institutional translation with several 4EU+ partners, and have been an advisory board member of the PIETRA project on the institutional translation practices of the Catholic Church. I am currently editing a special issue of The Interpreter and Translator Trainer, and co-editing another issue of Jurisprudence - Revue Critique.