 Your new post is loading...
|
Scooped by
Charles Tiayon
Today, 11:19 AM
|
As the availability and performance of AI for language editing and translation continues to improve, we can imagine a future in which everyone can use their own language to write, assess and read science. The question is, how can we achieve it? Tatsuya Amano, Lynne Bowker, Andrew Burton-Jones
In an ideal world, academic publishing is about “removing barriers and promoting inclusion in knowledge creation and sharing, and publishing research outputs that enable everyone to learn from, reuse and build upon scientific knowledge”. The use of English as the common language of science has boosted international scholarly communication, including publishing, but has also posed unignorable barriers to the progress and application of science. For example, reading, writing, and publishing papers requires much more time and effort for scientists whose first language is not English compared to native English speakers [1], which can lead to higher levels of anxiety and dissatisfaction [2]. Centralizing the publication of research around English also undermines the ability of people with limited English proficiency to read and use the research [3] and drives international research to ignore science published in other languages [4]. The scientific community urgently needs to move beyond the use of English as the singular default language to ensure that all scientists (and other actors and stakeholders) have an equal opportunity to access, contribute to, and benefit from science, regardless of their backgrounds [5].
A primary reasons that science has not yet become fully multilingual is that translation can be slow and costly. Artificial intelligence (AI), however, may finally allow us to overcome this problem, as it can provide useful, often free or affordable, support in language editing and translation [6]. Large language models are already widely used in academic writing, especially in countries where English is not widely spoken [7]. Existing AI has limitations, most notably variations in the availability and performance of AI among different languages [8]. However, assuming that this situation will continue to improve, we can now imagine two futures for academic publishing in which we could leverage the power of AI to overcome language barriers and improve equity in the publication, synthesis, and application of science.
In Future 1, English would continue to be the lingua franca in science (Fig 1A). Although international journals would continue to publish in English, researchers with limited English proficiency could write papers in their own language and use AI to translate them into English before submission. They could also use AI to translate English-language papers into their own language when reading, reviewing, and editing those papers. Scientific knowledge would continue to be centralized around English, but the use of AI would help to make science more easily producible and accessible for those with limited English proficiency.
Expand
Fig 1. Two futures for academic publishing using artificial intelligence language tools.
Information communicated in English is shown in pale blue and that in a language other than English (Japanese in this example) is shown in orange. (A) In Future 1, scientific papers continue to be published in English. Artificial intelligence (AI) is used by those with limited English proficiency to translate information between their preferred language and English when writing, assessing (reviewing and editing) and reading papers. (B) In Future 2, scientific papers are published in any language of the authors’ choice (English or Japanese in this example). AI is used by those without proficiency in the publication language (e.g., Japanese) to translate information between that language and their preferred language (e.g., English). Here, only one non-English language is shown for simplicity, but translation may be between different non-English languages. For example, Reviewer #1 may read a Japanese-written paper and provide feedback in Spanish, Reviewer #2 may do the same in Chinese, the Editor may read the paper and the reviewers’ comments and provide feedback in Arabic, and the authors would read all of their feedback in Japanese, all through AI translation. doi:10.1371/journal.pbio.3003215.g001
More »
This future is less ‘disruptive’ because the scientific community would continue to operate using the existing publishing system. Also, as most AI models are disproportionately trained on English-language data, translation to and from English tends to be of higher quality than translation between non-English languages. However, Future 1 would have various drawbacks. Inequality between fluent and non-fluent English speakers would remain; any negative consequences of using AI in academic publishing, including translation inaccuracies and the financial cost of using AI tools, would be imposed only on non-fluent English speakers. As long as scientific knowledge is centralized around English, the ongoing ‘domain loss’ (the idea that the growing use of English in a certain domain leads to other languages losing status and eventually not being used at all [9]) for other languages will not decelerate and could even intensify. New concepts in science may be described only in English, and other languages may not even have terms for new concepts. People will not be able to talk about science in their own language easily, and this may further isolate science from non-scientists, potentially leading to a lower uptake of science in decision-making and less trust in science among the general public.
Now imagine another future, Future 2, in which academic journals publish papers written in any language (Fig 1B). This would enable authors to write and submit papers in their own language. Here, assessors (editors and reviewers) and recipients of science (both scientists and non-scientists) would use AI to read those papers in their own language. A major advantage of this future would be that everyone can use their own language for science, which would help maintain and promote the diversity of science and scientific languages. This would be a giant leap forwards for 95% of the world’s population (native speakers of languages other than English) who, at present, have little choice but to conduct science in English. Publishing science in other languages could also help to halt domain loss and facilitate the understanding and use of science in countries where English is not widely spoken.
That said, making academic publishing multilingual in a fair way will not be easy. For example, even with AI tools, will people find and read English-language papers and papers in unfamiliar languages equally frequently? Will the evaluation of papers written in a non-English language be conducted in an unbiased manner? Given that AI translation is inevitably imperfect, especially for low-resource languages, this future could introduce another bias in the assessment, visibility and use of science depending on the language of publication. Various pragmatic roadblocks also exist; for instance, literature search systems would need to integrate multilingual metadata and cross-language information retrieval to allow users to search for literature written in different languages. The AI-driven automation of literature searching, screening and data extraction using multilingual models should help researchers to better use evidence in multiple languages [10]. However, encouraging publishing in languages beyond English will require a systemic change, as the current English-based assessment of science and scientists drives scientists to publish in English, even in countries where English is not widely spoken.
We should be mindful about the inaccuracy of AI translation and its consequences in science. But we also need to understand trade-offs between the consequences of AI inaccuracy and the considerable benefits of overcoming existing language barriers. The acceptable risk of using AI translation likely differs depending on purpose and discipline. Training subject area experts continues to be essential for spotting improbable AI translations, and the involvement of such experts will be necessary, especially where misunderstanding evidence can have serious implications.
Despite these and other counterarguments, Future 2 is still our preferred future, as it would truly democratize academic publishing. People often express concern that AI translation does not meet a phantom gold standard. But the reality is that issues around inaccurate use and understanding of English are already widespread in every process of academic publishing, and these are now entirely attributed to a lack of effort by researchers with lower English proficiency. The use of AI can at least create a more level playing field in the sense that it does not privilege one language above all others. There is no single big step towards Future 2; what we need are small stepping stones, such as experimenting with multilingual publication in just a few languages. To make a start, we have launched various initiatives [11] and encourage others to follow suit. AI will no doubt be integrated into all elements of the academic publishing workflow in the near future, and we believe now is the time for the scientific community to start discussing how we can use its benefits to move towards making science multilingual.
- Abstract
As the availability and performance of artificial intelligence for language editing and translation continues to improve, we can imagine a future in which everyone can use their own language to write, assess, and read science. The question is, how can we achieve it?
Citation: Amano T, Bowker L, Burton-Jones A (2025) AI-mediated translation presents two possible futures for academic publishing in a multilingual world. PLoS Biol 23(6): e3003215. doi:10.1371/journal.pbio.3003215 Published: June 23, 2025
Researchers across Africa, Asia and the Middle East are building their own language models designed for local tongues, cultural nuance and digital independence
"In a high-stakes artificial intelligence race between the United States and China, an equally transformative movement is taking shape elsewhere. From Cape Town to Bangalore, from Cairo to Riyadh, researchers, engineers and public institutions are building homegrown AI systems, models that speak not just in local languages, but with regional insight and cultural depth.
The dominant narrative in AI, particularly since the early 2020s, has focused on a handful of US-based companies like OpenAI with GPT, Google with Gemini, Meta’s LLaMa, Anthropic’s Claude. They vie to build ever larger and more capable models. Earlier in 2025, China’s DeepSeek, a Hangzhou-based startup, added a new twist by releasing large language models (LLMs) that rival their American counterparts, with a smaller computational demand. But increasingly, researchers across the Global South are challenging the notion that technological leadership in AI is the exclusive domain of these two superpowers.
Instead, scientists and institutions in countries like India, South Africa, Egypt and Saudi Arabia are rethinking the very premise of generative AI. Their focus is not on scaling up, but on scaling right, building models that work for local users, in their languages, and within their social and economic realities.
“How do we make sure that the entire planet benefits from AI?” asks Benjamin Rosman, a professor at the University of the Witwatersrand and a lead developer of InkubaLM, a generative model trained on five African languages. “I want more and more voices to be in the conversation”.
Beyond English, beyond Silicon Valley
Large language models work by training on massive troves of online text. While the latest versions of GPT, Gemini or LLaMa boast multilingual capabilities, the overwhelming presence of English-language material and Western cultural contexts in these datasets skews their outputs. For speakers of Hindi, Arabic, Swahili, Xhosa and countless other languages, that means AI systems may not only stumble over grammar and syntax, they can also miss the point entirely.
“In Indian languages, large models trained on English data just don’t perform well,” says Janki Nawale, a linguist at AI4Bharat, a lab at the Indian Institute of Technology Madras. “There are cultural nuances, dialectal variations, and even non-standard scripts that make translation and understanding difficult.” Nawale’s team builds supervised datasets and evaluation benchmarks for what specialists call “low resource” languages, those that lack robust digital corpora for machine learning.
It’s not just a question of grammar or vocabulary. “The meaning often lies in the implication,” says Vukosi Marivate, a professor of computer science at the University of Pretoria, in South Africa. “In isiXhosa, the words are one thing but what’s being implied is what really matters.” Marivate co-leads Masakhane NLP, a pan-African collective of AI researchers that recently developed AFROBENCH, a rigorous benchmark for evaluating how well large language models perform on 64 African languages across 15 tasks. The results, published in a preprint in March, revealed major gaps in performance between English and nearly all African languages, especially with open-source models.
Similar concerns arise in the Arabic-speaking world. “If English dominates the training process, the answers will be filtered through a Western lens rather than an Arab one,” says Mekki Habib, a robotics professor at the American University in Cairo. A 2024 preprint from the Tunisian AI firm Clusterlab finds that many multilingual models fail to capture Arabic’s syntactic complexity or cultural frames of reference, particularly in dialect-rich contexts.
Governments step in
For many countries in the Global South, the stakes are geopolitical as well as linguistic. Dependence on Western or Chinese AI infrastructure could mean diminished sovereignty over information, technology, and even national narratives. In response, governments are pouring resources into creating their own models.
Saudi Arabia’s national AI authority, SDAIA, has built ‘ALLaM,’ an Arabic-first model based on Meta’s LLaMa-2, enriched with more than 540 billion Arabic tokens. The United Arab Emirates has backed several initiatives, including ‘Jais,’ an open-source Arabic-English model built by MBZUAI in collaboration with US chipmaker Cerebras Systems and the Abu Dhabi firm Inception. Another UAE-backed project, Noor, focuses on educational and Islamic applications.
In Qatar, researchers at Hamad Bin Khalifa University, and the Qatar Computing Research Institute, have developed the Fanar platform and its LLMs Fanar Star and Fanar Prime. Trained on a trillion tokens of Arabic, English, and code, Fanar’s tokenization approach is specifically engineered to reflect Arabic’s rich morphology and syntax.
India has emerged as a major hub for AI localization. In 2024, the government launched BharatGen, a public-private initiative funded with 235 crore (€26 million) initiative aimed at building foundation models attuned to India’s vast linguistic and cultural diversity. The project is led by the Indian Institute of Technology in Bombay and also involves its sister organizations in Hyderabad, Mandi, Kanpur, Indore, and Madras. The programme’s first product, e-vikrAI, can generate product descriptions and pricing suggestions from images in various Indic languages. Startups like Ola-backed Krutrim and CoRover’s BharatGPT have jumped in, while Google’s Indian lab unveiled MuRIL, a language model trained exclusively on Indian languages. The Indian governments’ AI Mission has received more than180 proposals from local researchers and startups to build national-scale AI infrastructure and large language models, and the Bengaluru-based company, AI Sarvam, has been selected to build India’s first ‘sovereign’ LLM, expected to be fluent in various Indian languages.
In Africa, much of the energy comes from the ground up. Masakhane NLP and Deep Learning Indaba, a pan-African academic movement, have created a decentralized research culture across the continent. One notable offshoot, Johannesburg-based Lelapa AI, launched InkubaLM in September 2024. It’s a ‘small language model’ (SLM) focused on five African languages with broad reach: Swahili, Hausa, Yoruba, isiZulu and isiXhosa.
“With only 0.4 billion parameters, it performs comparably to much larger models,” says Rosman. The model’s compact size and efficiency are designed to meet Africa’s infrastructure constraints while serving real-world applications. Another African model is UlizaLlama, a 7-billion parameter model developed by the Kenyan foundation Jacaranda Health, to support new and expectant mothers with AI-driven support in Swahili, Hausa, Yoruba, Xhosa, and Zulu.
India’s research scene is similarly vibrant. The AI4Bharat laboratory at IIT Madras has just released IndicTrans2, that supports translation across all 22 scheduled Indian languages. Sarvam AI, another startup, released its first LLM last year to support 10 major Indian languages. And KissanAI, co-founded by Pratik Desai, develops generative AI tools to deliver agricultural advice to farmers in their native languages.
The data dilemma
Yet building LLMs for underrepresented languages poses enormous challenges. Chief among them is data scarcity. “Even Hindi datasets are tiny compared to English,” says Tapas Kumar Mishra, a professor at the National Institute of Technology, Rourkela in eastern India. “So, training models from scratch is unlikely to match English-based models in performance.”
Rosman agrees. “The big-data paradigm doesn’t work for African languages. We simply don’t have the volume.” His team is pioneering alternative approaches like the Esethu Framework, a protocol for ethically collecting speech datasets from native speakers and redistributing revenue back to further development of AI tools for under-resourced languages. The project’s pilot used read speech from isiXhosa speakers, complete with metadata, to build voice-based applications.
In Arab nations, similar work is underway. Clusterlab’s 101 Billion Arabic Words Dataset is the largest of its kind, meticulously extracted and cleaned from the web to support Arabic-first model training.
The cost of staying local
But for all the innovation, practical obstacles remain. “The return on investment is low,” says KissanAI’s Desai. “The market for regional language models is big, but those with purchasing power still work in English.” And while Western tech companies attract the best minds globally, including many Indian and African scientists, researchers at home often face limited funding, patchy computing infrastructure, and unclear legal frameworks around data and privacy.
“There’s still a lack of sustainable funding, a shortage of specialists, and insufficient integration with educational or public systems,” warns Habib, the Cairo-based professor. “All of this has to change.”
A different vision for AI
Despite the hurdles, what’s emerging is a distinct vision for AI in the Global South – one that favours practical impact over prestige, and community ownership over corporate secrecy.
“There’s more emphasis here on solving real problems for real people,” says Nawale of AI4Bharat. Rather than chasing benchmark scores, researchers are aiming for relevance: tools for farmers, students, and small business owners.
And openness matters. “Some companies claim to be open-source, but they only release the model weights, not the data,” Marivate says. “With InkubaLM, we release both. We want others to build on what we’ve done, to do it better.”
In a global contest often measured in teraflops and tokens, these efforts may seem modest. But for the billions who speak the world’s less-resourced languages, they represent a future in which AI doesn’t just speak to them, but with them."
Sibusiso Biyela, Amr Rageh and Shakoor Rather
20 May 2025
https://www.natureasia.com/en/nmiddleeast/article/10.1038/nmiddleeast.2025.65
#metaglossia_mundus
"Traduction, paix et confiance : JMT 2025
Le Conseil de la FIT a le plaisir d’annoncer le thème de la Journée mondiale de la traduction (JMT) 2025, inspiré par l’Année internationale de la paix et de la confiance des Nations Unies et par celui du Congrès de la FIT2025, Maîtres de la machine : façonnons l’intelligence de demain. Le thème de la JMT 2025 est
La traduction, garantie de votre confiance en l’avenir
Dans des temps troublés où souvent se joue l’avenir de la paix, où la défiance s’insinue dans les échanges internationaux, il met à l’honneur la confiance dans les relations humaines, spécifiquement le rôle des traducteurs et traductrices, interprètes et terminologues, garanties de fiabilité des communications, artisans du dialogue en confiance, maîtres des outils d’IA générative et de traduction automatique.
Dans sa résolution sur l’Année de la paix et de la confiance, l’Assemblée générale de l’ONU constate aussi la nécessité de prévenir et résoudre les conflits par le dialogue et la diplomatie. Si dans les négociations internationales, les interprètes ne sont pas visibles, nous savons que ces conversations stratégiques seraient impossibles sans leur travail et ne sauraient être laissées aux mains de la technologie.
Le thème rappelle également la résolution A/RES/71/288 de 2017 faisant du 30 septembre une Journée internationale du réseau des Nations Unies honorant la contribution de la traduction, l’interprétation et la terminologie professionnelles au rapprochement des nations et à la promotion de la paix et du développement.
Comme chaque année depuis plus de 35 ans, le Conseil encourage les membres à adopter pour leurs célébrations le thème de l’année. Et le Comité permanent des partenariats externes vous invite le 26 septembre 2025 à fêter la JMT en ligne avec la FIT lors du webinaire annuel."
https://fr.translatio.fit-ift.org/2025/03/28/traduction-paix-et-confiance-jmt-2025/
#metaglossia_mundus
"Ce webinaire est co-organisé par l'UNESCO et Translation Commons dans le cadre de la Décennie internationale des langues autĺochtones (DILA2022-2032) pour célébrer la Journée internationale de la traduction.
Traduction des langues autochtones : Façonner un avenir digne de confiance
30 septembre 2025 - 4:00 pm - 30 septembre 2025 - 6:30 pm
Location
UNESCO Headquarters, Paris, France
Rooms :
VIRTUAL ROOM A
Type :
Cat VII – Seminar and training
Arrangement type :
Virtual
Rejoignez l'UNESCO et Translation Commons pour un événement en ligne célébrant la Journée internationale de la traduction ! Le thème de cette année, la confiance dans la traduction autochtone, englobe plusieurs dimensions : la confiance dans l'exactitude et la sensibilité culturelle des traductions, la confiance dans le traitement éthique des connaissances et des données autochtones, et la confiance dans les technologies employées, en particulier en ce qui concerne l'intelligence artificielle. Il s'agit de s'assurer que le processus de traduction respecte l'autonomie des communautés autochtones. L'instauration de cette confiance nécessite la participation active des communautés autochtones, la transparence, des lignes directrices claires et le respect des droits moraux et matériels, tant pour les communautés que pour les professionnels de la langue concernés. L'établissement de la confiance est essentiel pour la préservation, la revitalisation et la promotion des langues autochtones, et pour garantir que leurs connaissances et leurs cultures sont représentées de manière précise et respectueuse.
Le webinaire sera en anglais.
Objectifs principaux
Le webinaire permettra de :
Explorer l'évolution des rôles et l'avenir des professionnels de la langue autochtone en examinant l'impact de la technologie, de l'IA et de l'évolution des compétences nécessaires dans le domaine.
Souligner l'importance de l'expertise humaine pour garantir l'exactitude, la sensibilité culturelle et la préservation de la signification spirituelle dans la traduction des langues autochtones.
Discuter et aborder les considérations éthiques liées à l'utilisation de l'IA et de la technologie numérique dans la traduction en langue autochtone, y compris la propriété des données, la protection de la vie privée, les préjugés et les cadres dirigés par la communauté.
Faciliter le dialogue sur les meilleures pratiques en matière de traduction et de validation communautaires, en garantissant une participation active et le respect de l'autonomie autochtone.
Sensibiliser aux droits moraux et matériels des communautés autochtones et des traducteurs/interprètes, en mettant l'accent sur une compensation, une reconnaissance et un soutien équitables.
Examiner le droit à l'accès culturel et linguistique pour les communautés autochtones, y compris l'aide juridique et l'interprétation dans des domaines critiques tels que les soins de santé.
Promouvoir la collaboration intergénérationnelle et les approches novatrices du travail linguistique à l'ère numérique.
Fournir des études de cas qui mettent en évidence le rôle essentiel des traducteurs humains dans la préservation du patrimoine linguistique et culturel.
Programme (en anglais seulement)
7.00 am PDT / 4 pm CEST
Welcome Address
7.10 am PDT / 4.10 pm CEST
The Enduring Trust in Human Voices: Ensuring Accuracy and Cultural Integrity within Indigenous Communities
7.50 am PST / 4.50 pm CEST
Navigating the Future: The Evolving Roles of Indigenous Language Professionals
8.30 am PST / 5.30 pm CEST
Ethical Pathways in the Digital Age: AI and the Future of Indigenous Language Translation
9.10 am PST / 6.10 pm CEST
Closing Address"
https://www.unesco.org/fr/articles/traduction-en-langues-autochtones-faconner-un-avenir-digne-de-confiance
#metaglossia_mundus
Traductrice, interprète, la Française Karine Martin accompagne au quotidien les expatriés francophones dans les démarches administratives de particuliers ou d'entreprises.
Karine Martin : "S’expatrier en Espagne demande beaucoup de traductions assermentées"
Pour les particuliers ou les entreprises, la traductrice, interprète, Karine Martin aide au quotidien les expatriés francophones dans les démarches administratives. Rencontre.
Écrit par Simon Legentil
Publié le 23 juin 2025, mis à jour le 25 juin 2025
Karine Martin pose ses valises en Espagne au début des années 2000, juste après ses études. Elle débute sa carrière à la Chambre de commerce, puis se lance dans l’entrepreneuriat en fondant Fidélité Idiomas, une entreprise de traduction et d’interprétation, à Alicante en 2006. Quelques années plus tard, elle s’installe à Barcelone pour son dynamisme, le cadre de vie et la proximité avec la France. Elle devient déléguée consulaire en 2021. Aujourd’hui, cette traductrice passionnée accompagne au quotidien la communauté francophone de Barcelone.
Fidélité Idiomas, concrètement, qu'est-ce que c'est ?
À la base, c'était moi en tant que traductrice-interprète. Espagnol, anglais vers le français. Au fil du temps, mes clients m'ont demandé de gérer des projets multilingues. Et c'est là où j'ai commencé à travailler en tant qu'agence avec des collègues freelance dans d'autres langues. Je n’ai pas de salariés, mais je collabore presque toujours avec les mêmes personnes. Vous savez, les bons traducteurs ne veulent pas être salariés. Ils préfèrent rester maîtres de leur temps, maîtres de leurs projets, maîtres de leurs clients.
Il y a deux versants de Fidélité Idiomas. Le premier, la traduction à l'écrit de textes simples, c'est-à-dire non assermentée. Ce sont des traductions pour des sites web, des annonces ou des textes commerciaux. Ensuite, il y a les traductions assermentées. Beaucoup de documents juridiques ont besoin d'être assermentés pour être présentés légalement devant les administrations. C’est le cas quand des Français viennent s’installer ici par exemple.
L’autre volet, c'est l’interprétation. J'accompagne des chefs d'entreprise qui viennent faire des réunions, des négociations ou qui viennent voir leurs équipes. Et il y a l'interprétation simultanée, pour des congrès ou des visites avec des audioguides par exemple.
On traduit énormément pour les Français qui viennent s'installer en Espagne.
Et comment les entreprises ou les particuliers font appel à vous ? C'est sur votre site ou c'est vous qui les démarchez ?
J'ai un site web. Je fais partie de nombreux réseaux d'entrepreneurs à Barcelone comme La Peña ou la Chambre de commerce française. Et je suis répertoriée sur la liste des interprètes recommandés par le consulat. Au niveau de l'interprétariat, les entreprises font leurs recherches et je pense que je remonte dans les recherches internet.
Quelles sont les plus grosses demandes des francophones qui vous sollicitent ?
Alors, on traduit énormément pour les Français qui viennent s'installer en Espagne. Il faut beaucoup de traductions assermentées pour s'expatrier ici. Je pense que les plus fortes demandes que nous avons sont les livrets de famille, les actes de naissance ou des diplômes. Ça va être des traductions assermentées pour les particuliers. Et ensuite, pour les entreprises, on va traduire beaucoup de documents destinés à ouvrir des filiales ici, comme le Kbis. C’est, on va dire, la carte d'identité de l'entreprise en France. Et ensuite, des contrats de collaboration, des négociations, etc. Ça, ce sont des traductions assermentées aussi.
En termes de délai, quand une entreprise ou un particulier demande de traduire un numéro ou une carte d'identité de l'entreprise, combien de temps ça met ?
Tout dépend des documents et de l’urgence de ceux-ci. Des documents courts, en général d'une à trois pages, on est très réactif. C'est d'ailleurs une des caractéristiques que nous avons et c'est pour cela qu'on est très connu de la communauté pour la gestion des urgences. Étant moi passée par ces démarches-là, je connais le stress que représente le fait de rater son rendez-vous parce qu'il nous manque un papier. Il nous arrive, de l'après-midi pour le lendemain, d'avoir la traduction.
Maintenant, pour des entreprises, les documents sont plus longs. Un document d’une vingtaine de pages, il faut quand même un minimum de temps pour pouvoir procéder. Mais on est toujours attentif et en bonne collaboration avec nos clients.
Le traducteur engage sa responsabilité sur les traductions, l'IA n'a rien d'officiel.
Comment devient-on traducteur-interprète ?
Alors, traducteur-interprète, ce sont de vraies études ! C'est bien que vous me posiez la question. On devient traducteur en faisant des études de traducteur. En France, il y a des écoles de traduction. En Espagne, cela se fait à l'université surtout.
C’est un vrai métier. De nos jours, l’IA vient réellement nous concurrencer. Alors, il faut la regarder avec un œil expert. Parce que sur de la traduction simple, comme des mails, ou des textes n’ayant rien d’engageant, je dois reconnaître que l'IA est très forte et rapide. Donc là, je me tire un peu une balle dans le pied. Mais il faut être réaliste. Maintenant, quand on parle des contrats, sur des documents confidentiels ou personnels, je pense qu'il faut être très précautionneux au niveau de la confidentialité. Il y a aussi un gros risque de contresens. Et je le répète, dans tout ce qui est juridique, on reste des experts, on engage notre responsabilité sur les traductions, l'IA n'a rien d'officiel."
https://lepetitjournal.com/barcelone/installation/karine-martin-expatriation-espagne-traductions-assermentees-416055
#metaglossia_mundus
Le ministère de l'Europe et des Affaires étrangères a lancé récemment DiploIA, un dispositif de traduction et de transcription multilingue à destination de ses quelque 13 000 agents, dont une part exerce à l'étranger. Pour répondre aux besoins des missions sensibles, le ministère a dû articuler sécurité et technologie.
"Avec DiploIA, le Quai d’Orsay met la traduction et la transcription au centre de sa stratégie IA
Par Victoria Beurnez - 23 juin 2025 -
Le ministère de l'Europe et des Affaires étrangères a lancé récemment DiploIA, un dispositif de traduction et de transcription multilingue à destination de ses quelque 13 000 agents, dont une part exerce à l'étranger. Pour répondre aux besoins des missions sensibles, le ministère a dû articuler sécurité et technologie.
Comme quelques-uns avant lui, le ministère de l’Europe et des Affaires étrangères s’est désormais lancé dans le déploiement d’outils d’intelligence artificielle internes, à destination de ses agents. Contrairement à ce qui se fait ces derniers mois, il n’est pas question ici d’un chatbot, mais d’outils pensés précisément pour les besoins des agents, notamment en dehors de la France. C’est en réfléchissant au plus près de ces besoins que la direction du numérique du ministère a déployé DiploIA auprès de ses quelque 13 000 agents, en France et à l’étranger, depuis le mois de mai.
“Nous avons une vraie attente de la part de nos agents sur de tels outils”, explique Virginie Rozière, directrice du numérique au ministère de l’Europe et des Affaires étrangères, auprès d’Acteurs publics. Pour autant, il n’était pas question de se lancer dans des outils sans réflexion en amont, qui pourraient dès lors “relever du gagdet”, tempère la directrice..."
https://acteurspublics.fr/articles/le-quai-dorsay-deploie-de-lia-au-service-de-ses-agents/
#metaglossia_mundus
The Faculty of Translation and Interpreting (FTI, formerly ETI) is deeply committed to fundamental and applied research. Our areas of expertise are: translation studies, interpretation, terminology and machine-assisted translation, multilingual communication and international mediation.
"Research at the FTI
At the Faculty of Translation and Interpreting (FTI) we are committed to innovative research at the highest levels of excellence. Our areas of expertise include translation studies, interpreting studies, terminology and machine-assisted translation, multilingual communication and international mediation.
Our researchers are key players in major national and international projects in these domains and work with likeminded partners across the globe. Their research output is disseminated through high quality journals and other specialised publications, including Parallèles, the FTI’s own journal.
Research by department:
The Department of Translation
The Department of Translation Technology
The Interpreting Department
Research groups:
Centre for Legal and Institutional Translation Studies (Transius)
Economics, Languages and Education Research Group (ELF)
Laboratory for Research in Interpreting and Complex Language Processing (LaborInt)
Interpreting and Technology (InTTech)
Access through interpretation-mediated communication (AXS)"
https://www.unige.ch/fti/en/recherches
#metaglossia_mundus
BEING one of the most multicultural regional centres in the country, Greater Shepparton has many bilingual people stepping up to be unofficial interpreters and translators for family, friends and co-workers.These community members help with everyday day scenarios, but when they are called to interpret educational, financial, legal and health matters, it becomes extremely challenging and could go against personal boundaries while breaching ethical codes.
"Interpreting the needs of multicultural communities
June 25, 2025
BEING one of the most multicultural regional centres in the country, Greater Shepparton has many bilingual people stepping up to be unofficial interpreters and translators for family, friends and co-workers.
These community members help with everyday day scenarios, but when they are called to interpret educational, financial, legal and health matters, it becomes extremely challenging and could go against personal boundaries while breaching ethical codes.
To help local organisations navigate the complexities of interpreting and translating, Monash University in conjunction with GSCC’s Resilience in Recovery team hosted several workshops for local stakeholders. The aim was to understand the role and responsibilities of interpreters and the importance of the National Accreditation Authority for Translators and Interpreters (NAATI) certification.
“The job of the interpreter is to convey the meaning and linguistic features from one language to another. That is law. It’s not to convey emotion, not to convey opinion, not to convey perception, it’s just to convey the meaning,” said Dr Leah Gerber, senior lecturer of translation and interpreting studies at Monash University.
INTERPRETERS AND TRANSLATORS… Monash University in conjunction with GSCC’s Resilience in Recovery team hosted workshops to help local organisations navigate the complexities of interpreting and translating language for local multicultural communities. Pictured is Dr Leah Gerber, senior lecturer of translation and interpreting studies at Monash University. Photo: Aaron Cordy
Linguistic skills, even fluency do not give you the skills and understanding of the ethics required to be an interpreter.
“Speaking the languages, reading the languages, writing the languages, does not mean you are a translator or an interpreter. In Australia, you usually must complete a one-and-a-half to two-year Master’s program or Advanced Diploma in order to qualify to sit the NAATI test,” said Dr Gerber.
The training sessions provided by Monash and the GSCC flood recovery team play an important role in this space, but more work needs to be done to help our local multicultural community access the training and services of qualified interpreters and translators.
“I think regional communities are really disadvantaged in the sense that all training for translators and interpreters happens in metropolitan Melbourne. There aren’t any regional opportunities other than either going to Melbourne to study or accessing the courses online, which doesn’t suit everybody’s learning needs,” said Dr Gerber.
“I think there’s a lack of awareness also of the very specific challenges that regional communities face, whether it’s in terms of a weather event, whether it’s in terms of particular language demands that you can’t just get an interpreter from Melbourne to come out to Shepparton when you need them. Those kinds of things are important to recognise.”"
https://www.sheppadviser.com.au/interpreting-the-needs-of-multicultural-communities/
#metaglossia_mundus
This study compares the quality of English-to-Arabic translations produced by Google Translate (GT) with those generated by student translators.
"Man vs. Machine: Can AI Outperform Student Translations?
Anas Alkhofi*
King Faisal University, Al-Ahsa, Saudi Arabia
This study compares the quality of English-to-Arabic translations produced by Google Translate (GT) with those generated by student translators. Despite advancements in neural machine translation technology, educators often remain skeptical about the reliability of AI tools like GT and often discourage their use. To investigate this perception, 20 Saudi university students majoring in English and Translation produced human translations in Arabic. These studentgenerated translations, along with their GT equivalents, were rated by 22 professors with experience in language-related fields. The analysis revealed a significant preference for GT translations over those produced by students, suggesting that GT's quality may exceed that of student translators. Interestingly, while GT translations were consistently rated higher, instructors often misattributed the better translations to students and the poorer ones to GT. This reveals a strong perceptual bias against AI-generated translations. The findings support the inclusion of AI-assisted translation tools in translation training. Incorporating these tools will help students prepare for a job market where AI is playing an increasingly important role.Incorporating such tools will help students prepare for a growing job market in which AI is playing a growing role. At the same time, educators should adopt strategies incorporating AI tools without sacrificing the development of students' core translation skills."
https://www.frontiersin.org/journals/artificial-intelligence/articles/10.3389/frai.2025.1624754/abstract
#metaglossia_mundus
Award, to be shared between poet and translator, is a joint project by three publishers and will give a $5,000 advance for a new collection
New prize for translated poetry aims to tap into boom for international-language writing
Award, to be shared between poet and translator, is a joint project by three publishers and will give a $5,000 advance for a new collection
Ella Creamer
Tue 24 Jun 2025 15.00 BST
A new poetry prize for collections translated into English is opening for entries next month.
Publishers Fitzcarraldo Editions, Giramondo Publishing and New Directions have launched the biennial Poetry in Translation prize, which will award an advance of $5,000 (£3,700) to be shared equally between poet and translator.
The winning collection will be published in the UK and Ireland by Fitzcarraldo Editions, in Australia and New Zealand by Giramondo and in North America by New Directions.
“We wanted to open our doors to new poetry in translation to give space and gain exposure to poetries we may not be aware of,” said Fitzcarraldo poetry editor Rachael Allen. “There is no other prize like this that we know.”
The prize announcement comes amid a sales boom in translated fiction in the UK. Joely Day, Allen’s co-editor at Fitzcarraldo, believes that “the space the work of translators has opened up in the reading lives of English speakers through the success of fiction in translation will also extend to poetry”.
Translated work is a focus of the three publishers behind the prize. Fitzcarraldo has published translated works by Nobel prize winners Olga Tokarczuk, Jon Fosse and Annie Ernaux. “Our prose lists have always maintained a roughly equitable balance between English-language and translation, and some of our greatest successes have been books in translation,” said Day. “We’d like to bring the same diversity of voices to our poetry publishing, and this prize will be a step in that direction.”
skip past newsletter promotion
Sign up to Bookmarks
Free weekly newsletter
Discover new books and learn more about your favourite authors with our expert reviews, interviews and news stories. Literary delights delivered direct to you
Enter your email address
Sign up
Privacy Notice: Newsletters may contain info about charities, online ads, and content funded by outside parties. For more information see our Privacy Policy. We use Google reCaptcha to protect our website and the Google Privacy Policy and Terms of Service apply.
after newsletter promotion
Four Nobels and counting: Fitzcarraldo, the little publisher that could
Read more
The prize is open to living poets from around the world, writing in any language other than English.
The prize is being launched to find works “which are formally innovative, which feel new, which have a strong and distinctive voice, which surprise and energise and move us,” said Day. “My personal hope is that the prize reaches fledgling or aspiring translators and provides an opening for them, that it enables translators of poetry in particular to find a platform and encourages translators who want to work with poetry to do so.”
Submissions will be open from 15 July to 15 August. A shortlist will be announced later this year, with the winner announced in January 2026 and publication of the winning collection scheduled for 2027.
The “unique” award “brings poetry from around the world into English, and foregrounds the essential role of translation in our literature,” said Nick Tapper, associate publisher at Giramondo. “Its global outlook will bring new readers to poets whose work deserves wide and sustained attention.”
This article was amended on 24 June 2025. An earlier version incorrectly listed the prize money amount as $3,000 rather than $5,000 in the subheading."
https://www.theguardian.com/books/2025/jun/24/new-prize-for-translated-poetry-aims-to-tap-into-boom-for-international-language-writing
#metaglossia_mundus
https://www.theguardian.com/books/2025/jun/24/new-prize-for-translated-poetry-aims-to-tap-into-boom-for-international-language-writing
#metaglossia_mundus
Microsoft has unveiled its on-device small language model, Mu, that allows users to change settings through natural language queries.
"Microsoft introduces small language model Mu to change settings in Windows 11
Accessible to Windows Insiders in the Dev Channel with Copilot+PCs, Mu responds to natural language queries
Microsoft on Monday (June 23, 2025) unveiled its on-device small language model, Mu, that allows users to change settings through natural language queries. The company said Windows Insiders in the Dev Channel with Copilot+PCs can access Mu.
The AI agent for Settings in Windows 11 was included in the existing search box for a seamless user experience, the company said.
“Mu is fully offloaded onto the Neural Processing Unit (NPU) and responds at over 100 tokens per second, meeting the demanding UX requirements of the agent in Settings scenario,” Vivek Pradeep, VP, Distinguished Engineer at Windows Applied Sciences noted in a company blog.
Trained over multiple phases using Nvidia’s A100 GPUs on Azure Machine Learning, Mu followed a similar technique as Microsoft’s previous small language model family, Phi.
It was pre-trained on “hundreds of billions of the highest-quality educational tokens, to learn language syntax, grammar, semantics and some world knowledge.” Mu was then distilled from the Phi models to enhance accuracy.
Published - June 24, 2025 02:03 pm IST"
Updated - June 24, 2025 02:44 pm IST
https://www.thehindu.com/sci-tech/technology/microsoft-introduces-small-language-model-mu-to-change-settings-in-windows-11/article69731231.ece
#metaglossia_mundus
"A tale of trauma and the perils of misinterpretation
The debut novel by the Albanian American writer Ledia Xhoga tackles past trauma and communication breakdowns
Past trauma haunts the debut novel by Albanian American writer Ledia Xhoga. When the narrator, an unnamed Albanian interpreter, agrees to accompany Alfred, a Kosovar torture survivor, to therapy sessions, her life begins to unravel.
The interpreter lives in Brooklyn with her husband, Billy, a film professor. When she invites Leyla, a Kurdish poet seeking asylum, and her friend to stay for the weekend, Billy’s unexpected return exposes the cracks in their marriage. His angry reaction seems disproportionate until it becomes clear that the narrator’s compulsion to help fellow immigrants is taking a toll on their relationship. We are given a hint of her psychological entanglement when she says: “Sometimes words affected me physically, causing as much nausea as motion sickness.” This manifests in her work with Alfred, prompting his therapist to fire her.
As pressure builds, Billy accepts a six-month artist’s residency in Hungary, while the narrator visits Albania. The novel takes on a meditative tone as we learn of her mother’s inability to leave her home and the narrator’s troubled childhood. Xhoga raises the tension back in Brooklyn, as her protagonist faces the fallout from helping Leyla escape a stalker employed by her abusive ex-husband.
Xhoga writes perceptively about the alienation of immigrants – the narrator straddles two worlds, adrift in both – and the disconnect between privilege and precarity. Of Billy’s friend, a professional violinist, she observes: “How could Anna, who was born and raised in the West Village … understand water or electricity shortages, or helping out ageing parents and relatives who had worked all their lives only to end up with a retirement that didn’t even cover their basic necessities?”
Part of the pleasure of the novel, which is related in a direct, matter-of-fact tone, comes from second-guessing what the (occasionally unreliable) narrator has withheld or misinterpreted and how it affects others. Alfred misreads her compassion. She misconstrues others’ expectations, risking her marriage and mental health. The book is a nuanced exploration of communication failures, blurred boundaries and the emotional cost of unchecked altruism.
Misinterpretation by Ledia Xhoga is published by Daunt (£10.99)"
Lucy Popescu
22 June 2025
https://observer.co.uk/culture/books/article/the-perils-of-misinterpretation
#metaglossia_mundus
"ABSTRACT: Language is an underrecognized social driver of health that disproportionately affects Hispanic/Latino (H/L) patients and can negatively impact comprehension and engagement with medical care. Professional interpretation services can help overcome language barriers in the clinic or hospital setting; however, interpreters are not universally available and, even when they are, the quality of communication can be limited. In this secondary analysis of a qualitative study which examined the barriers and facilitators H/L patients and their caregivers experience when navigating rectal cancer care, we sought to explore how Spanish-speaking rectal cancer patients and their caregivers communicated with oncology providers and the ways in which professional interpretation may have facilitated or detracted from their care experiences. Methods We conducted a community-partnered qualitative study to explore H/L patients with rectal cancer and their caregivers’ experiences with medical interpretation during their oncologic care. We developed an interview guide based on the Ecological Model of Health Behavior and iteratively refined it with input from our Community Advisory Board. Data analysis utilized grounded theory and reflexive thematic analysis to identify core themes. Results Over a 6-month period, we conducted 21 semi-structured interviews. Three major themes related to language arose from our review of coded transcripts: (1) interpreters’ use of medical jargon; (2) dialect discordance between patient and hospital interpreter; and (3) lack of trust in the interpretation process. Conclusion Our sample of H/L rectal cancer patients and their caregivers reported barriers due to overuse of medical jargon and dialectic differences, both of which eroded trust in the interpretation process. Beyond simply hiring more interpreters, hospitals would benefit from re-envisioning the patient-interpreter relationship as a means of improving communication and fostering trust in the healthcare system."
June 2025
Supportive Care in Cancer
33(7)
DOI: 10.1007/s00520-025-09657-6
Eleanor Brown
Min Young Kim
Zaria N. Cosby
et al.
Aaron J. Dawe
https://www.researchgate.net/publication/392797572_Interpreters_don%27t_tell_you_everything_experiences_with_medical_interpretation_among_Latino_cancer_patients_and_caregivers_with_limited_English_proficiency
#metaglossia_mundus
"Intercultural Preparation in Higher Education
ByNing Chen
Edition1st Edition
First Published 2025
eBook Published 27 August 2025
Pub. LocationLondon
Imprint Routledge
Pages112
eBook ISBN9781003661818
SubjectsEducation
ABSTRACT
Intercultural Preparation in Higher Education critically examines the complexities of intercultural preparation, challenging reductionist, competence-based approaches and emphasising critical reflection, linguistic sensitivity and intersectional perspectives, which offer transformative insight.
This concise yet comprehensive book draws on the author's extensive experience in preparing students and educators for diversity and intercultural engagement within Chinese and Finnish higher education – two distinct yet interconnected settings. Chen advocates for reflexive, context-sensitive frameworks that address power dynamics, ideological influences and epistemic inequalities shaping intercultural encounters. Synthesising global and Chinese research, the author deconstructs interculturality as an ideological construct and proposes expanding it to encompass diverse diversities. Through case studies of Chinese Minzu education, Finnish teacher training and China's mental health and well-being education, the author explores how to equip educators and students for glocalised (global + local) realities. The book argues that mental health education in China, for instance, can redefine interculturality beyond 'international' or 'ethnic' binaries. By interrogating dominant paradigms and proposing equitable alternatives, the author advances a balanced and inclusive vision of preparation for all in higher education and beyond.
The title is designed for educators, researchers, students and practitioners interested in intercultural studies, higher education, as well as internationalisation and globalisation."
https://www.taylorfrancis.com/books/mono/10.4324/9781003661818/intercultural-preparation-higher-education-ning-chen
#metaglossia_mundus
"By Santa Barbara Unified School District
Tue Jun 24, 2025 | 11:43am
Students in Santa Barbara High School’s groundbreaking Translation & Interpretation Pathway graduated with professional certificates that qualify them to enter the growing field of interpretation and translation. This innovative Career Technical Education (CTE) program celebrates the bilingual assets of Santa Barbara’s many native Spanish speakers and empowers students to transform their language skills into viable career opportunities. Santa Barbara High School’s Translation & Interpretation Pathway, led by teacher Alison Mendoza, stands as the first and only high school CTE pathway of its kind in the state of California.
“Developing this pathway has been a career highlight”, says Alison Mendoza. “By committing to offering a Translation & Interpretation program, we have reimagined Spanish course offerings and connected language learning to the world that surrounds students in a new way. Students receive Language Other Than English (UC A-G) credit for participating and put their language abilities into practice. Watching this year’s cohort leave their high school experience with various college options, a tangible career possibility, a deeper sense of who they are, and pride in their linguistic backgrounds aligns with why I wanted to become a teacher. The collaboration with our local community partners, the district LAU team, and higher education gives students opportunities that they often hadn’t imagined yet”.
This year, the program significantly expanded its offerings to include “The Community Interpreter®” course and certificate. To earn this industry-recognized credential, students completed a rigorous 40-hour training led by Sofia Rubalcava of the Santa Barbara Unified School District’s highly regarded Language Access Unit. Programs like this are designed to strengthen the connection between classroom learning and real-world careers by integrating industry training, all while allowing students to embrace and connect with their home languages and cultures.
“This expansion of the Translation & Interpretation Pathway is a testament to our commitment to providing students with tangible skills that directly lead to meaningful careers,” said Superintendent Hilda Maldonado. “By empowering our bilingual students with professional certificates, we are not only opening doors to vital professions but also reinforcing the value of their linguistic and cultural heritage within our community and beyond. It is a testament to the resiliency of our teachers and staff. They started this even though we were in a pandemic. At SB Unified, we are resilient, strong, and bold!”
The Translation & Interpretation Pathway equips students with foundational skills in translation and interpretation, emphasizing cultural competence, ethics, and the critical importance of language access in diverse community settings. Students apply their learning through real-world experiences, serving their local community, and engaging with professionals currently working in the fields of court, medical, and community interpreting. This year alone, students in the pathway have interpreted the daily morning announcements at their school, created student-friendly translations with the Santa Barbara Museum of Art, volunteered at the Unity Shoppe, established internship opportunities with the Santa Barbara County Courthouse, and provided crucial support to parents and the community at the Know Your Rights event alongside the Language Access Unit team.
Jose Navarrete, a court interpreter at the Santa Barbara County Courthouse, is one of many professionals who actively support the program. He underscores the critical need to inspire a new generation of interpreters. “You need to bring up a whole new generation of interpreters. Approximately 35% of interpreters in the state are 65 years old or older. So they’re going to retire pretty soon. We’re going to need a new generation,” said Navarrete.
The program’s impact is evident in the students themselves. Rubiell Angel Fernandez, a graduating pathway senior, shared, “I’ve had Ms. Mendoza all four years of high school. It really means a lot, seeing how much the program and the pathway have grown throughout the years.” Student Jay Valencia added, “It’s a rigorous program, but even still, it’s served us well, it’s going to give us options for the future.”
This expansion solidifies Santa Barbara High School’s Translation & Interpretation Pathway as a national model for preparing bilingual students for in-demand careers while fostering cultural understanding and community engagement."
https://www.independent.com/2025/06/24/santa-barbara-high-schools-translation-interpretation-pathway-expands-to-provide-professional-certificate/
#metaglossia_mundus
Translation expertise significantly influences how translators manage cognitive resources, yet the specific ways in which professional and novice translators differ in their metacognitive strategy use remain incompletely understood. This gap is particularly evident in specialised contexts such as academic translation, where complex terminology and intricate syntactic structures pose unique cognitive challenges. This study investigated how professional and student translators deploy metacognitive strategies when translating academic texts from Chinese to English, focusing on differences in cognitive resource allocation across translation stages. The study compared 30 professional translators and 30 graduate students in translation studies as they translated two academic article introductions. Using keystroke logging, we recorded detailed temporal data about participants’ translation processes, including thinking time, writing time, and resource consultation patterns. Quantitative analysis revealed that professional translators demonstrated significantly shorter thinking times compared to students, while maintaining higher writing-to-thinking ratios. Analysis of cognitive resource distribution showed that professionals maintained more consistent engagement across translation stages compared to students’ more variable patterns. These findings suggest that translation expertise manifests not merely in faster processing times, but in qualitatively different approaches to cognitive resource allocation. Professional translators demonstrated greater flexibility in switching between automatic and controlled processes, adapting their strategies according to textual challenges. These results refine our understanding of translation expertise while offering practical implications for translator training, suggesting specific ways to help novice translators develop more efficient cognitive strategies. The study contributes to translation process research by providing empirical evidence of how expertise shapes metacognitive strategy use in specialised translation contexts.
"Interpreters of more than 130 languages have been called into New Zealand courts over the past decade - and the costs have soared by 229%.
The Waikato Times took a deep dive into the issue in the wake of remarks from Judge David Cameron about a case with a Filipino interpreter.
Thanks to an Official Information Act request with the Ministry of Justice, the Waikato Times can reveal the increased cost of court interpretation, the complete list of languages and the most - and least - common.
In 2014, it cost the Ministry $1,751,874 for interpreter services, a figure that rose to $5,619,896 in 2023.
The Ministry - and several law and linguistic experts the Waikato Times spoke to - said however that the spike in costs does not necessarily equate to more foreign criminals in New Zealand...
The Ministry said the cost increase between 2022 and 2023 - from $3,475,847 to $5,619,896 - was due to factors including the Coronial inquiry into the Christchurch terror attacks.
They also cited a 2023 change “to increase the fees payable to interpreters”.
Professor Tafaoimalo Tologata Leilani Tuala-Warren, Dean of Te Piringa Faculty of Law at the University of Waikato, also cautioned that there could be a few explanations.
“We may be seeing more witnesses, for both prosecutions and defence, in criminal trials needing interpreters. It could also suggest that the cost charged by interpreters has increased. It is important to not make assumptions without the benefit of examining the disaggregation of the data.”
Tineke Jannick, who is conducting doctoral research into court interpreting in New Zealand at the Victoria University of Wellington, told the Waikato Times the language list may simply be a reflection of “the reality of the highly multi-cultural and multi-lingual nation we live in”.
“Over a quarter of New Zealand’s population are born overseas and there are more than 150 languages spoken here,” she said...
Jannick also said interpretation services were “vital to our justice system, particularly as they pertain to fair trial rights”.
One defence lawyer who spoke to the Waikato Times on condition of anonymity cited the 2022 Accredited Employer Work Visa (AEWV) for the spike in costs.
More than 150,000 visas had been approved since - most before increased English language requirements imposed in 2024, they said, and the scheme was “widely rorted”.
“Many migrants have paid money to agents to get a work visa and a job, only to discover when they arrive in New Zealand that the job offers very few hours, or is entirely fake.
“Some ... turn to crime to pay the bills. Others came to New Zealand with the express purpose of committing crime, knowing from the beginning that their job was fake, but using it to obtain a visa in circumstances where applications were being rubber-stamped by Immigration New Zealand in the post-Covid immigration boom,” they said.
“I have personally dealt with many people on AEWVs who turned to crime shortly after arriving on their visa.”
The data also revealed the five most translated languages over the past decade: Mandarin, Samoan, Punjabi, Tongan and Vietnamese.
Those least required are Bulgarian, Hiligaynon [a regional Filipino language], Lingala [a language spoken in the Democratic Republic of the Congo], Ndebele [South Africa and Zimbabwe] and Oromo [Ethiopia].
Dr Simon Overall, University of Otago linguistics postgraduate coordinator, said the top five list was interesting.
“Two of them, Samoan and Tongan, have relatively small numbers of speakers, but of course significant numbers of those speakers live in New Zealand.
“Of the five, all except Vietnamese appear in the list of the top 20 languages spoken in New Zealand (according to 2018 census data).”
“I don’t know what demographic reason there might be for Vietnamese being there – perhaps the Vietnamese speaking community in New Zealand has increased in recent years?”
“All in all, the lists give a good picture of the New Zealand linguistic situation and to me it really highlights the vital importance of interpreters and translators, since we can’t simply expect that everybody (whether New Zealand citizens or not) knows English well enough to be able to represent themselves in court proceedings.”
Benn Bathgate
benn.bathgate@stuff.co.nz
June 23, 2025
https://www.waikatotimes.co.nz/nz-news/360732341/afar-zulu-130-plus-languages-required-translation-court-and-cost
#metaglossia_mundus
"Imagine microscopes, thermometers, balances, pipettes and bunsen burners having native names, and science and all subjects and courses, taught in our own languages.
We would have been able to compete favourably and equitably within the so-called global village.
The Ministry of Education should boldly prioritise Ghanaian languages in our education system.
By doing so, we can create a more inclusive and effective learning environment for our students.
When students are taught in their native language, they are more likely to understand and retain complex concepts, leading to better academic performance.
Moreover, prioritising Ghanaian languages can help preserve our cultural heritage and promote national identity.
Our languages are an integral part of our history and traditions, and by promoting them, we can ensure their survival for future generations.
The current system, which prioritises foreign languages, particularly English, puts an undue burden on students, especially those from rural areas where English may not be widely spoken.
This can lead to a high dropout rate and poor academic performance.
By adopting a more localised approach to education, we can empower our students to succeed and contribute to the development of our nation.
It's time for the Ministry of Education to rethink our language policy and prioritise Ghanaian languages in our education system.
This can be achieved by developing curriculum materials in Ghanaian languages, training teachers to teach in local languages and encouraging the use of Ghanaian languages in schools and communities.
By taking these steps, we can create a more inclusive and effective educational system that benefits all students, regardless of their background or location.
It's time to put our languages and cultures at the forefront of our education system and give our students the tools they need to succeed.
Moses Sackie Agbemava,
Presbyterian Church of Ghana,
Dansoman Estates, Accra"
https://www.graphic.com.gh/features/opinion/ghana-news-ministry-of-education-prioritise-ghanaian-languages-over-any-foreign-language.html
Des centaines de personnes pratiquent le métier d'interprètes dans le réseau de la santé.
"Les interprètes, le « lien de confiance » pour les allophones dans le réseau de la santé
Accéder à la section Commentaires
L'interprète Olga Lacasse et la travailleuse sociale Geneviève Thibault discute avec une personne assise dos à la caméra.
Ouvrir en mode plein écran
L'interprète Olga Lacasse et la travailleuse sociale Geneviève Thibault ont accompagné de nombreux patients ensemble.
Flavie Sauvageau
Publié à 9 h 00 UTC+1
La version audio de cet article est générée par la synthèse vocale, une technologie basée sur l’intelligence artificielle.
Des centaines de personnes au Québec pratiquent le métier d’interprètes dans le milieu de la santé. Ils côtoient au quotidien des patients qui ne parlent pas encore le français et permettent de faciliter leur lien avec les soignants.
Depuis 2022, Olga Lacasse pratique le métier d'interprète dans le réseau de la santé, à Québec. À la demande des professionnels de la santé, elle accompagne les patients qui ne parlent ou ne comprennent pas encore le français à leurs rendez-vous médicaux ou avec les services sociaux.
Moi, je traduis en ukrainien et j'ai commencé à cause de la guerre en Ukraine. Avec l'arrivée d'Ukrainiens qui ne parlaient pas français au début, il y avait un besoin urgent d’interprètes, se souvient celle qui travaillait auparavant en relations internationales.
Suivis médicaux, rendez-vous d’urgence, rencontres avec la DPJ ou encore accouchements… Elle est appelée à travailler dans toutes sortes de situations, pas toujours faciles à entendre, notamment lorsqu’elle aide les déplacés d’une guerre.
Un cahier posé sur des genoux. Une main qui tient un crayon pour écrire dans le cahier.
Ouvrir en mode plein écran
Les interprètes répondent en moyenne à 270 demandes par jour dans le réseau de la santé.
AILLEURS SUR INFO : La confiance envers le ministre Drainville est entamée, disent la FAE et la CSQ
En tant qu’humaine, on ne peut pas rester toujours sans émotion, mais on essaye toujours de rester neutre, empathique, mais neutre pour s'assurer que les messages de médecins ou de notre client soient bien transmis, explique-t-elle, précisant qu’elle travaille toujours pour les deux parties : le médecin et le patient.
Publicité
Olga fait au moins 10 accompagnements par semaine. Elle rappelle que son rôle n’est pas de faire le travail des médecins, mais plutôt de les aider à transmettre leur message aux usagers, avec respect et de manière neutre.
Des services disponibles en 90 langues
On a des interprètes pour couvrir 90 langues, donc c'est quand même énorme, lance Stéphanie Fiset, adjointe à la direction de l'Hôpital Jeffery Hale - Saint Brigid's.
À travers toute la province, ils sont 400 interprètes à répondre à environ 270 demandes d’interprétation ou de traduction, chaque jour, indique celle qui est aussi gestionnaire de la banque d’interprètes. Au total, 80 000 demandes sont faites dans la province par année.
Et le recrutement est constant. Il s’adapte aux différentes vagues migratoires selon les régions.
Les qualités qu'on recherche? Un niveau qui est quand même assez élevé. [Il] faut vraiment être capable d'être en mesure de comprendre les nuances. Quand on parle de langage médical aussi, énumère la gestionnaire.
On a besoin des gens qui sont capables d'avoir une coupure professionnelle, parce que parfois ils entendent des choses qui ne sont vraiment pas faciles. Il faut être capable d'être humain, mais de rester quand même distant, croit-elle.
Les cinq langues les plus demandées pour les services d’interprétation
Langue
Nombre de demandes en 2024-2025
Proportion par rapport au nombre total de demandes
Espagnol 7250 36 %
Arabe 3150 16%
Swahili 1870 9 %
Népalais 1575 8 %
Rohingya 832 4 %
À elles seules, ces cinq langues représentent 73 % des demandes totales au CIUSSS de la Capitale-Nationale.
Source : CIUSSS de la Capitale-Nationale
Des « courroies de transmission »
La présence d’interprètes comme Olga Lacasse lors de rendez-vous fait vraiment la différence pour les patients, constate Geneviève Thibault, qui est travailleuse sociale à la Clinique santé des réfugiés et aux Services aux demandeurs d’asile de l’hôpital Jeffery-Hale.
Moi je suis vraiment là pour voir comment se passe l'adaptation et l'intégration au Québec, donc je suis là aussi pour évaluer, voir au niveau des chocs post-traumatiques, est-ce que ça se passe bien? La personne, au niveau psychosocial et psychologique, comment elle va? , décrit-elle.
L'interprète Olga Lacasse et la travailleuse sociale Geneviève Thibault.
Ouvrir en mode plein écran
L'interprète Olga Lacasse et la travailleuse sociale Geneviève Thibault.
Photo : Radio-Canada / Bruno Giguère
Ces rencontres sont souvent des moments très émotifs entre l’intervenante et son patient, même si la barrière de la langue se dresse entre eux.
Publicité
D'où l'importance que les interprètes soient présents, souligne-t-elle, parce que je ne parle pas le swahili, je ne parle pas l'ukrainien, je ne parle pas le kinyarwanda. Donc eux sont vraiment notre courroie de transmission.
Et leur apport a un effet visible pour les patients dans son bureau. Pour eux, c'est le lien de confiance. C'est quelqu'un qui, souvent, partage la même culture. C'est quelqu'un qui a les mêmes référents culturels, fait que je le vois clairement, décrit celle qui a souvent travaillé avec Olga Lacasse.
Un de leurs souvenirs marquants ensemble : un suivi de grossesse. Je me rappelle qu'on a les petits yeux mouillés parce qu’on entend le cœur fœtal. Puis là, la maman est émue, puis là, Olga est émue. Puis là, je suis émue, raconte-t-elle avec le sourire.
Fusion des banques d'interprètes du Québec
L’automne dernier, la banque d’interprète du CIUSSS de la Capitale-Nationale a fusionné avec celle du CIUSSS du Centre-Sud-de-l 'Île-de-Montréal, pour créer la nouvelle Banque d'interprètes du réseau de la santé et des services sociaux.
L’objectif, c'est vraiment d'avoir un service 24/7 pour les personnes du Québec qui ne sont pas capables de s'exprimer en français, explique Stéphanie Fiset.
Le fait d’offrir des soins de santé dans une autre langue que le français ou l’anglais a été codifié dans la réforme de la Charte de la langue française."
https://ici.radio-canada.ca/nouvelle/2174202/interprete-soins-de-sante-allophone-quebec
#metaglossia_mundus
The newly established Hellenic Language Institute (INEL) aims to position Greek as a global language of thought and cultural connection. At its inaugural conference at the Acropolis Museum, experts from 15 countries explored ways to preserve, teach, and revitalise Greek for future generations.
"...Greek is not a relic of the past. It’s a living, dynamic language with global potential—and a vital tool for the future. That’s the message at the heart of the new Hellenic Language Institute (INEL), led by Evangelia Georgantzi, a veteran educator with decades of experience teaching Greek as a second and foreign language.
Speaking to AMNA from the Acropolis Museum, where INEL’s inaugural conference recently took place, Georgantzi laid out the institute’s mission: to preserve, promote, and expand the use of Greek worldwide, not only among diaspora communities but for anyone seeking access to a language rich in philosophy, precision, and cultural depth.
A language of thought, not just tradition
Founded with support from the Qualco Foundation and Qualco Group, INEL envisions Greek not simply as a ‘heritage language’ but as a contemporary tool of expression, cognition, and intercultural understanding. An estimated 13–15 million people worldwide currently speak Greek, either as a first or second language. Georgantzi’s goal is to grow that number through education, cultural diplomacy, and digital innovation.
“The Greek language has an unmatched ability to express abstract concepts with philosophical depth,” she explained. “It offers both historical continuity and a modern voice, bridging the past and pointing toward the future.”
The challenge for the Diaspora: A living legacy
Georgantzi pointed to a dual challenge, especially for Greek communities abroad: preventing language loss among second, third, and now fourth-generation Greek Australians, while also making the language appealing to broader audiences...
Modern tools for a timeless language
INEL’s strategy includes initiatives like the Ellinika Gia Sas (“Greek for You”) program, supported in 11 languages, and the embrace of digital tools that support hybrid and personalised learning.
At the conference, over 20 universities from 15 countries shared innovative teaching methods. Notably, Australia’s use of the VoiceThread platform and the unveiling of Kri-Kri, a Greek AI language tool developed by the Athena Research Center as a companion to ChatGPT, drew major interest.
Sessions focused on experiential learning, intercultural mediation, and AI in education. “The exchange of best practices can benefit both Greek education abroad and domestic public schools,” Georgantzi added.
The future: A permanent global forum
A major outcome of the conference is INEL’s plan to institutionalise it as a recurring global forum. The next steps include creating a scientific network and building partnerships with international organisations and Greek language programs abroad..."
23 June 2025 3:29pm
Twitter: @NeosKosmos
https://neoskosmos.com/en/2025/06/23/news/diaspora/greek-is-not-just-heritage-its-a-global-language-for-the-future-says-hellenic-language-institute-president/
#metaglossia_mundus
This collection of essays explores the translation of Petrarch’s vernacular verse (Canzoniere and Triumphi) in early modern Britain, from the first Tudor translations to its many literary transformations and cultural re-appropriations in the Elizabethan and Stuart periods.
"Translating Petrarch in early modern Britain gathers twelve essays by international scholars focusing on the translation of Petrarch’s vernacular verse (Canzoniere and Triumphi) into English, from the Tudor age to the mid-seventeenth century (and beyond).
Approaching translation as an interpretive process, but also a mode of literary emulation and cultural engagement with Petrarch’s prestigious precedent, the collection explores the complex and interconnected trajectories of both poetic works in English and Scottish literary milieux. While situating each translation in its distinct historical, material, and literary context, the essays trace the reception of Petrarch’s works in early modern Britain through the combined processes of linguistic and metric innovation, literary imitation, musical adaptation and cultural and material ‘domestication’.
The collection sheds light on the origins and development of early modern English Petrarchism as part of wider transnational – and indeed, translational — European literary culture."
Manchester University Press
Edited by Marie-Alice Belle, Riccardo Raimondo, Francesco Venturi
9781526173034
1 September 2025
Hardback
384pp
240x170mm
RFB
AUD$210.00
NZD$240.00
https://newsouthbooks.com.au/books/translating-petrarch-in-early-modern-britain/#:~:text=The%20collection%20sheds%20light%20on,%2C%20translational%20%E2%80%94%20European%20literary%20culture.
#metaglossia_mundus
Book Awards India: Writers express concern as JCB Prize for Literature quietly halts operations, ending a key platform for Indian languages.
Malayalam writer Benyamin, who won the debut award in 2018 for his book “Jasmine Days” translated into English by Shahnaz Habib, said that the news is “deeply disappointing”. “...as the JCB Prize was a highly promising recognition for Indian literature -- not just Indian English writing, but also regional languages. It was a respected and influential award that played a significant role in helping Indian literature reach a global audience,” the author of “Goat Days” told PTI.
In its seven editions, the award has gone to translated works of fiction five times. In 2020, S Hareesh’s “Moustache”, translated from the Malayalam by Jayashree Kalathil, won the award. M Mukundan’s “Delhi: A Soliloquy”, translated from the Malayamal by Fathima E V and Nandakumar K, won the prize in 2021. The award in 2022 went to Urdu author Khalid Jawed’s “The Paradise of Food”, translated by Baran Farooqui, followed by Tamil author Perumal Murugan’s “Fire Bird”, translated by Janani Kannan, in 2023. “This is a great loss for Indian languages. Malayalam translations have won the award three times, Tamil once, and Urdu has also been recognized. It's truly disheartening,” Murugan said.
The award in 2019 was given to Madhuri Vijay for “The Far Field”. Author Namita Gokhale said that it is possible that they might be “reconsidering the format”, while appreciating the work done by the JCB Prize...
Karthika VK, publisher of Westland Books, said that the prize added a “great deal to the publishing ecosystem”. "I am sorry to hear that it's shutting down. We can only hope others will come come along to fill the gap and sponsor prizes, grants, residencies, to support writers and translators,” she said.
In its run of seven years, the award courted controversy ahead of the prize announcement last year when over a hundred writers, poets and publishers came out with an open letter condemning JCB, the British bulldozer manufacturer and the organiser of the literature prize, for allegedly “uprooting” the lives of poor and marginalised in India as well as Palestine.
Benyamin said that the criticism “missed the point”. “...it’s not the equipment that should be blamed, but those who choose to misuse it. The same machines are also used to build also. I can't align myself with that kind of reasoning, especially when the prize itself had such a positive impact. I hope they will reconsider and they will reinstate the prize for Indian literature,” he said. The award followed the process of call for entry, a longlist of 10 books, a shortlist of five books and finally, a winner. Each of the shortlisted authors was awarded Rs one lakh, and if the shortlisted piece were a translation, the translator received Rs 50,000. If a translated work won the prize, the author would take home Rs 25 lakh while the translator received the prize money of Rs 10 lakh.
https://www.deccanherald.com/india/jcb-prize-for-literature-makes-quiet-exit-causes-concern-among-writers-3596542
#metaglossia_mundus
Deepa Bhasthi’s translator’s note for Heart Lamp, winner of the International Booker Prize 2025, challenges translation conventions. As translators claim creative agency and push back against conventions like italicisation, the way a translator’s role is perceived is evolving.
"‘Against Italics’ - Deepa Bhasthi’s translator’s note for Heart Lamp, which won the International Booker Prize 2025, made headlines. The translator’s note section, which is usually skipped, drew attention this time. “Setting aside the futile debate of what is lost and found in translation”, Ms. Bhasthi talks about her deliberate choice to not use italics for the Kannada, Urdu, and Arabic words that remain untranslated in English.
Ms. Bhasthi ends the translator’s note saying, “Italics serve to not only distract visually, but more importantly, they announce words as imported from another language, exoticising them and keeping them alien to English. By not italicising them, I hope the reader can come to these words without interference, and in the process of reading with the flow, perhaps even learn a new word or two in another language. Same goes for footnotes – there are none”. While talking to The Hindu, Ms. Bhasthi says this translator’s note has been met with a lot of love and care, and attention by readers as well."
https://www.thehindu.com/education/beyond-heart-lamps-success-translators-are-slowly-getting-the-credit-thats-due/article69716831.ece
#metaglossia_mundus
"Cassava announces another MoU – to train LLMs in African languages and ‘context’
Cassava Technologies is building an African AI ecosystem, from its own AI factory to alliances with NVIDIA, developers and a solutions firm
Cassava Technologies has signed a Memorandum of Understanding (MoU) with a startup Vambo AI to build “Africa’s own large language models” – that is, LLMs trained using African languages and culture. Vambo AI will be given access to Cassava’s AI Factory to train and deploy a series of open-source language models.
“We’re not just building tools, we’re building the future,” said Vambo AI’s CEO, Chido Dzinotyiwei.
This is just the latest tie-up in a recent flurry for Cassava, having teamed up with Zindi in May to court AI developers, and, inevitably, NVIDIA, the world’s biggest AI chip maker. As TechZIM put it, added to the deal with Vambo AI, this constitutes “a full-stack African AI ecosystem”. Then’s there’s also the recent tie-up with Sand Technologies, the a global enterprise AI solutions company.
It launched its own AI unit in July 2024. In March it announced “new strategic appointments to the leadership team as the organisation ushers in the next growth phase and ensures future competitiveness”. The new appointments were announced after the company closed several rounds of funding and investment, “underpinned by a strong balance sheet”.
Vambo AI’s multilingual AI tools
Vambo is building multilingual AI tools, for translation, transcription, content generation and smart search across 44 African and 20 international languages. Its imminent Jua Series of open-source models will be released under permissive licence which will allow local developers and researchers to build on top of it, without the bother of non-disclosure agreements or gatekeeping."
Annie Turner
16 June 2025
https://www.mobileeurope.co.uk/cassava-announces-another-mou-to-train-llms-in-african-languages-and-context/
#metaglossia_mundus
Salaire divisé par deux, moins de contrats, perte de sens… Les traducteurs percutés de plein fouet par l’IA générative
Publié le 22 juin 2025 à 09:31
Depuis deux ans, leur vie professionnelle a été bouleversée par la vague de l’IA générative. Récit du combat donquichottesque des traducteurs qui refusent de voir leurs métiers se paupériser, et qui dénoncent la réalité de l’impact de l’IA sur leur secteur, et plus largement sur la langue, la pensée, la culture et la société
« Toutes les semaines, on a des collègues (traducteurs) qui arrêtent. C’est l’hécatombe ». En novembre 2022, ChatGPT, l’agent conversationnel d’OpenAI, était lancé aux États-Unis, marquant le début de la vague de l’intelligence artificielle (IA). Si plusieurs études ont prédit la fin de certains métiers, le tsunami a bel et bien déferlé sur le secteur des traducteurs. Et près de trois ans plus tard : les professionnels de la traduction que nous avons interrogés sont unanimes. Il y a bien eu un avant et un après ChatGPT.
« Ces deux dernières années, la dégradation de nos conditions de travail s’est vraiment accélérée, notamment pour des personnes (indépendantes) comme moi qui travaillent pour des intermédiaires, comme les agences de traduction », confie Julie, qui a souhaité rester anonyme. « On n’en est plus à des commandes de traduction, mais exclusivement à de la post-édition », déplore cette indépendante qui traduit depuis plusieurs années des textes techniques et marketing.
Depuis deux ans, leur vie professionnelle a été bouleversée par la vague de l'IA générative. Récit du combat donquichottesque des traducteurs qui refusent de voir leurs métiers se paupériser, et qui dénoncent la réalité de l'impact de l'IA sur leur secteur, et plus largement sur la langue, la pensée, la culture et la société.
"Pour les néophytes, « la “post-édition”, c’est le mot qu’on utilisait avant pour l’IA générative », précise-t-elle : il s’agit de travailler non plus exclusivement sur le texte original, mais sur une première version « traduite » par des outils de traduction automatique (TAN) ou d’IA générative (IAG). « Les agences de traduction cherchent depuis longtemps à nous imposer l’utilisation de ces systèmes. Et jusqu’en 2023, on pouvait encore dire non. Mais depuis ChatGPT, c’est terminé », regrette la trentenaire.
Une rémunération divisée par deux
« On a vraiment vu un basculement du discours du : “c’est inévitable, maintenant, vous utilisez la traduction automatique/l’IA générative, et vous serez payés deux fois moins” », confirme Laura, une autre indépendante, membre du collectif de traducteurs IA-lerte générale. Résultat : « J’ai un salaire divisé par deux. Certains clients m’ont remplacé par Deepl » (une plateforme de traduction automatique grand public NDLR), souligne celle qui traduit des textes techniques et marketing. La tendance se constaterait dans tous les métiers de la traduction, bien que certains domaines soient plus touchés que d’autres.
Un sondage réalisé en mars 2025 par l’ATAA, l’association des traducteurs et adaptateurs de l’audiovisuel, et auquel ont répondu près de 450 traducteurs, montre une baisse d’activité particulièrement marquée entre le premier trimestre 2024 et le début de cette année. Parmi toutes les activités du secteur (doublage, sous-titrage, voice-over, audiodescription, adaptation de scénario et de jeux vidéo), ce sont la voice-over et les sous-titrages pour sourds et malentendants qui sont de plus en plus traduits automatiquement ou via l’IAG. Les premières digues à sauter avant les autres ?
« On a demandé à une confrère, qui s’était occupée d’adapter en français une série étrangère, de s’occuper de la nouvelle saison, mais pour la première fois, avec de la post-édition. Elle a dit non, je ne fais pas de post édition. On lui a répondu : entendu, tu n’adapteras plus cette série », rapporte Stéphanie Lenoir, traductrice, adaptatrice, et présidente de l’ATAA.
Une autre étude, menée cette fois en avril 2025 par l’ATLF, l’association des traducteurs littéraires de France sur les conditions de travail des professionnels du secteur, et qui paraîtra début juillet, montre une réelle baisse de la rémunération des traducteurs d’édition, « oscillant entre 30 et 64 % », nous détaille l’organisation.
« C’est comme si une personne qui n’y connait rien en plomberie imposait une clé aux plombiers »
Pour Laura, « il y a vraiment une bataille qui a été perdue ou évitée, on ne sait pas trop. Mais nous, on n’a pas souscrit à ça, on n’est pas d’accord avec le fait d’avoir été mis devant le fait accompli » – comprenez : être obligé de passer par de la post-édition, et donc par de l’IA générative (IAG), pour traduire un texte. Face à ces remous, certains se sont regroupés en collectifs comme IA-lerte générale, ou En chair et en os. Leur objectif est de sensibiliser le grand public à la réalité de l’impact de l’IA sur leurs métiers, mais aussi à la réalité du travail de traduction.
Il faut dire que dans ce domaine, les malentendus sont légion, soulignent unanimement les traductrices que nous avons interrogées. À commencer par l’idée même de la post-édition. Son postulat est de dire : ces outils d’IAG et de traduction automatique vont faire gagner du temps aux traducteurs en « prémâchant » un premier jet, détaille Julie. Le problème est que « ces outils nous sont imposés dans 99 % des cas par des gens qui ne sont pas traducteurs, et qui vont dire : “tu as ChatGPT, c’est super” », poursuit-elle. Pour Valentine, une autre professionnelle indépendante qui traduit deux langues étrangères vers le français depuis une dizaine d’années, « c’est comme si une personne qui n’y connait rien en plomberie imposait une clé à molette aux plombiers ».
Il faut dire qu’au premier abord, « si on prend un texte et qu’on le met dans l’IA, il en ressort des phrases qui sont, on va dire, grammaticales », reconnait-elle. « Ce n’est pas du charabia. Les gens sont impressionnés, ils se disent : j’ai appuyé sur “entrée”, j’ai ma traduction, alors à quoi ça sert de demander à quelqu’un de le faire, et de le payer », poursuit Julie.
« Mais la traduction, ce n’est pas cela. Ca se joue au niveau du sens, et le sens ne se réduit pas aux mots séparément », ajoute-t-elle. Les mots ne sont pas des choses bien distinctes et découpées. Un mot est un infini de possibilités, qui dépend de plusieurs facteurs culturels, émotionnels, inconscients, interpersonnels. Ce que fait l’IA, c’est mettre des mots les uns à la suite des autres. Et comme c’est probabiliste, de temps en temps, ça fonctionne, mais seulement en surface », estime-t-elle.
Même son de cloche chez Stéphanie Lenoir, traductrice dans l’audiovisuel et adaptatrice, pour qui : « les outils d’intelligence artificielle et de traduction automatique considèrent la langue comme un système de codage de la pensée. Traduire avec l’IAG, c’est passer d’un code A à un code B. Mais on ne peut pas réduire un texte à un code », tacle-t-elle.
« À la fin, on se dégrade »
Or, si les traductrices interrogées prennent le temps à nous décrire ce qu’est une traduction (humaine), c’est pour nous expliquer à quel point le fait de se voir imposer des outils de traduction automatique ou d’IA générative affecte leur travail en lui-même. Imaginez : « j’ai le texte source (dans sa langue d’origine) et j’ai la traduction automatique », expose Julie. Le problème est que « les LLM ne sont pas capables de faire toute l’analyse du texte. Il y a beaucoup de contresens, des mots pour lesquels le LLM va choisir le mauvais sens », sans compter « les répétitions, et la lourdeur de style », liste-t-elle.
Dit autrement, « je me retrouve à corriger un texte qui est bourré d’erreurs que moi, je n’aurais pas fait en amont, parce que je sais qu’il y a plein de pièges à éviter. Mais cela, la machine ne le sait pas. Et c’est extrêmement fatigant de faire cela. Je ne peux même pas dire que ça équivaut à reprendre le travail d’un débutant, parce que même un traducteur débutant ne ferait pas des telles erreurs », poursuit-elle.
« Pour en avoir discuté avec d’autres confrères et consœurs, reprendre de la post-édition, ça nous prend autant de temps, voir plus, que de traduire un texte de zéro. On est immédiatement pollué par les suggestions de la traduction automatique. Et c’est fatigant de passer son temps à se détacher de cette première impression qui est truffée de fautes et d’erreurs », regrette Julie.
Selon l’étude de l’ATLF à paraître en juillet prochain, les outils d’IAG et de traduction automatique sont peu utilisés aujourd’hui par les traducteurs littéraires. 93 % des professionnels interrogés déclaraient ne pas utiliser de tels systèmes pour traduire leur texte.
Mais certains comme Valentine y ont été économiquement contraints. La trentenaire explique avoir passé plusieurs semaines à faire de la post-édition « pas bien payée, où il faut aller vite ». Son constat : à la fin, « j’avais perdu toute notion de qualité, j’avais perdu la boussole à l’intérieur qui me disait : est-ce un bon texte ou pas ? Je me suis dit : “il faut que j’arrête tout de suite”. On dégrade nos compétences en fait, même sans le vouloir. On se dégrade. Et on n’arrive plus à écrire après », déplore-t-elle.
Si l’indépendante a aujourd’hui des commandes qui lui permettent de se passer de post-édition, elle ne sait pas combien de temps cela va durer. « Si j’étais à fond sur mes principes, je n’aurais pas de travail », reconnaît-elle. Pour Laura qui a fait le choix de refuser toute post-édition, les commandes sont devenues rares. « Si je n’étais pas en couple, je serais déjà aller travailler au Mc Do », confie-t-elle, avant d’ajouter : « beaucoup de gens n’ont pas le choix, ceux qui ont des enfants ou qui sont célibataires acceptent d’être payés deux fois moins pour régler les factures, et manger ».
Problème de sécurité, vol, travailleurs du clic…
Pour les traductrices que nous avons interrogées, la dégradation de la qualité des textes, de leur travail et de leur salaire est loin d’être le seul problème. D’abord, des textes particulièrement sensibles ou confidentiels peuvent finir dans la moulinette des outils d’IA, à l’image des notices d’utilisation de médicaments, de défibrillateur ou du nucléaire. « On risque d’avoir des erreurs grossières qui peuvent être graves. À côté du risque de se ridiculiser pour les textes marketing, pour tout ce qui est médical, nucléaire, très technique, la vie des gens est en jeu. On attend quoi, qu’il y ait un drame pour que les gens commencent à réagir ? », s’insurge Laura. Certains professionnels interrogés, engagés par des agences de traduction, se demandent même si leur client final sait réellement que leur texte est passé par de la « post-édition ».
Autre problème de taille : tout le système repose sur « du vol, des conséquences désastreuses pour la planète, et des travailleurs du clic » invisibilisés, déplore Laura, qui s’interroge : « finalement ça ne dérange pas trop les gens qu’on nous ait volé notre travail ». Les outils d’IA générative ont été formés sur des textes protégés par le droit d’auteur, sans l’autorisation de leurs auteurs et ayants-droit.
Les auteurs peinent à opposer leur refus d’utilisation de leurs œuvres dans ces phases d’entrainement. « Quand on pense qu’avant, on menaçait les adolescents qui téléchargeaient des films », lance Valentine, aussi membre du collectif AI-lerte générale, en référence à Hadopi qui envoyait des courriers et prononçait parfois des amendes en cas de téléchargement illégal.
Mais pour Stéphanie Lenoir qui considère les traducteurs comme des lanceurs d’alerte, touchés en première ligne par l’IA générative, le problème est bien plus profond. « Ce n’est pas tellement parce que nos métiers vont être les premiers à disparaître. Dans notre secteur, on est confrontés à ces sujets depuis cinq, voire dix ans. On a du recul et on est capables de bien comprendre les dérives que ça entraîne. Et le risque essentiel contre lequel il faut que les pouvoirs publics et le grand public se mobilisent, c’est la standardisation de notre langue à vitesse grand V, la normalisation de la pensée et de tous les modes d’expression ».
Car si notre système de langue est passé à une moulinette de traduction automatique (et de l’IAG), « on se remettra toujours dans le mode de fonctionnement de la pensée du modèle majoritaire (le modèle anglo-saxon NDLR) », développe-t-elle. « Ça ne permet plus d’exprimer de nuances, de positions divergentes. (On le voit avec) l’algorithme, fait pour fonctionner sur l’occurrence la plus fréquente. Donc ça ferme le champ de découverte, ça ferme le champ d’expression, ça ferme le champ de réflexion », martèle-t-elle.
À l’université, « on apprend déjà que le futur de la traduction, c’est la post-édition. En sortant de la fac, les jeunes ne font que ça. Donc en fait, ils ne savent pas traduire. Ce qui m’effraie, c’est qu’un jour, plus personne ne saura traduire, écrire par soi-même, penser par soi-même », souffle Valentine. La bataille n’est pour autant pas encore perdue, estime Laura. « Si j’ai un message à adresser aux traducteurs, qui sont dans la galère et qui ont envie de tout abandonner, c’est : gardez espoir. Plus on va être nombreux à sensibiliser, à défendre notre métier, à dénoncer, et plus on aura un impact. La traduction est un métier qui existe depuis des millénaires. Ce n’est pas possible qu’elle disparaisse comme ça, du jour au lendemain »."
https://www.01net.com/actualites/salaire-divise-par-deux-moins-de-contrats-perte-de-sens-les-traducteurs-percutes-de-plein-fouet-par-lia-generative.html
#metaglossia_mundus
|
As the availability and performance of AI for language editing and translation continues to improve, we can imagine a future in which everyone can use their own language to write, assess and read science. The question is, how can we achieve it?
Tatsuya Amano, Lynne Bowker, Andrew Burton-Jones
In an ideal world, academic publishing is about “removing barriers and promoting inclusion in knowledge creation and sharing, and publishing research outputs that enable everyone to learn from, reuse and build upon scientific knowledge”. The use of English as the common language of science has boosted international scholarly communication, including publishing, but has also posed unignorable barriers to the progress and application of science. For example, reading, writing, and publishing papers requires much more time and effort for scientists whose first language is not English compared to native English speakers [1], which can lead to higher levels of anxiety and dissatisfaction [2]. Centralizing the publication of research around English also undermines the ability of people with limited English proficiency to read and use the research [3] and drives international research to ignore science published in other languages [4]. The scientific community urgently needs to move beyond the use of English as the singular default language to ensure that all scientists (and other actors and stakeholders) have an equal opportunity to access, contribute to, and benefit from science, regardless of their backgrounds [5].
A primary reasons that science has not yet become fully multilingual is that translation can be slow and costly. Artificial intelligence (AI), however, may finally allow us to overcome this problem, as it can provide useful, often free or affordable, support in language editing and translation [6]. Large language models are already widely used in academic writing, especially in countries where English is not widely spoken [7]. Existing AI has limitations, most notably variations in the availability and performance of AI among different languages [8]. However, assuming that this situation will continue to improve, we can now imagine two futures for academic publishing in which we could leverage the power of AI to overcome language barriers and improve equity in the publication, synthesis, and application of science.
In Future 1, English would continue to be the lingua franca in science (Fig 1A). Although international journals would continue to publish in English, researchers with limited English proficiency could write papers in their own language and use AI to translate them into English before submission. They could also use AI to translate English-language papers into their own language when reading, reviewing, and editing those papers. Scientific knowledge would continue to be centralized around English, but the use of AI would help to make science more easily producible and accessible for those with limited English proficiency.
Expand
Fig 1. Two futures for academic publishing using artificial intelligence language tools.
Information communicated in English is shown in pale blue and that in a language other than English (Japanese in this example) is shown in orange. (A) In Future 1, scientific papers continue to be published in English. Artificial intelligence (AI) is used by those with limited English proficiency to translate information between their preferred language and English when writing, assessing (reviewing and editing) and reading papers. (B) In Future 2, scientific papers are published in any language of the authors’ choice (English or Japanese in this example). AI is used by those without proficiency in the publication language (e.g., Japanese) to translate information between that language and their preferred language (e.g., English). Here, only one non-English language is shown for simplicity, but translation may be between different non-English languages. For example, Reviewer #1 may read a Japanese-written paper and provide feedback in Spanish, Reviewer #2 may do the same in Chinese, the Editor may read the paper and the reviewers’ comments and provide feedback in Arabic, and the authors would read all of their feedback in Japanese, all through AI translation.
doi:10.1371/journal.pbio.3003215.g001
More »
This future is less ‘disruptive’ because the scientific community would continue to operate using the existing publishing system. Also, as most AI models are disproportionately trained on English-language data, translation to and from English tends to be of higher quality than translation between non-English languages. However, Future 1 would have various drawbacks. Inequality between fluent and non-fluent English speakers would remain; any negative consequences of using AI in academic publishing, including translation inaccuracies and the financial cost of using AI tools, would be imposed only on non-fluent English speakers. As long as scientific knowledge is centralized around English, the ongoing ‘domain loss’ (the idea that the growing use of English in a certain domain leads to other languages losing status and eventually not being used at all [9]) for other languages will not decelerate and could even intensify. New concepts in science may be described only in English, and other languages may not even have terms for new concepts. People will not be able to talk about science in their own language easily, and this may further isolate science from non-scientists, potentially leading to a lower uptake of science in decision-making and less trust in science among the general public.
Now imagine another future, Future 2, in which academic journals publish papers written in any language (Fig 1B). This would enable authors to write and submit papers in their own language. Here, assessors (editors and reviewers) and recipients of science (both scientists and non-scientists) would use AI to read those papers in their own language. A major advantage of this future would be that everyone can use their own language for science, which would help maintain and promote the diversity of science and scientific languages. This would be a giant leap forwards for 95% of the world’s population (native speakers of languages other than English) who, at present, have little choice but to conduct science in English. Publishing science in other languages could also help to halt domain loss and facilitate the understanding and use of science in countries where English is not widely spoken.
That said, making academic publishing multilingual in a fair way will not be easy. For example, even with AI tools, will people find and read English-language papers and papers in unfamiliar languages equally frequently? Will the evaluation of papers written in a non-English language be conducted in an unbiased manner? Given that AI translation is inevitably imperfect, especially for low-resource languages, this future could introduce another bias in the assessment, visibility and use of science depending on the language of publication. Various pragmatic roadblocks also exist; for instance, literature search systems would need to integrate multilingual metadata and cross-language information retrieval to allow users to search for literature written in different languages. The AI-driven automation of literature searching, screening and data extraction using multilingual models should help researchers to better use evidence in multiple languages [10]. However, encouraging publishing in languages beyond English will require a systemic change, as the current English-based assessment of science and scientists drives scientists to publish in English, even in countries where English is not widely spoken.
We should be mindful about the inaccuracy of AI translation and its consequences in science. But we also need to understand trade-offs between the consequences of AI inaccuracy and the considerable benefits of overcoming existing language barriers. The acceptable risk of using AI translation likely differs depending on purpose and discipline. Training subject area experts continues to be essential for spotting improbable AI translations, and the involvement of such experts will be necessary, especially where misunderstanding evidence can have serious implications.
Despite these and other counterarguments, Future 2 is still our preferred future, as it would truly democratize academic publishing. People often express concern that AI translation does not meet a phantom gold standard. But the reality is that issues around inaccurate use and understanding of English are already widespread in every process of academic publishing, and these are now entirely attributed to a lack of effort by researchers with lower English proficiency. The use of AI can at least create a more level playing field in the sense that it does not privilege one language above all others. There is no single big step towards Future 2; what we need are small stepping stones, such as experimenting with multilingual publication in just a few languages. To make a start, we have launched various initiatives [11] and encourage others to follow suit. AI will no doubt be integrated into all elements of the academic publishing workflow in the near future, and we believe now is the time for the scientific community to start discussing how we can use its benefits to move towards making science multilingual.
Abstract
As the availability and performance of artificial intelligence for language editing and translation continues to improve, we can imagine a future in which everyone can use their own language to write, assess, and read science. The question is, how can we achieve it?
Citation: Amano T, Bowker L, Burton-Jones A (2025) AI-mediated translation presents two possible futures for academic publishing in a multilingual world. PLoS Biol 23(6): e3003215. doi:10.1371/journal.pbio.3003215
Published: June 23, 2025