 Your new post is loading...
|
Scooped by
Charles Tiayon
December 29, 2019 2:09 AM
|
Chinese technology giant Baidu has overtaken Google and Microsoft in an artificial intelligence competition designed to test how well a machine can understand human language. Baidu, which is often referred to as China‘s Google, achieved the highest ever score in the General Language Understanding Evaluation (Glue) – widely considered to be the benchmark for AI language
"Charles Tiayon, cartographer of the translation world
Charles Tiayon is a translator, lecturer and consultant whose Scoop.it activity reflects the professional universe named in his bio: translation, terminology, lexicography and intercultural issues. Through METAGLOSSIA, his public selection tracks language as a field of work, a field of rights, and a field being altered by AI.
A systems view of language
Tiayon looks at translation as a living system. His selections connect interpreters, translators, dictionary makers, universities, publishers, language activists and software companies, rather than treating them as separate worlds.
Technology with memory
AI is a recurring concern, but it is not treated as the whole story. Real-time speech translation, OCR and voice AI appear alongside older and slower forms of linguistic work: dictionaries, terminology, literary translation and language education.
A multilingual compass
The geographic spread is notably international, with attention to French, Spanish, Chinese, Arabic, Maltese, Breton, Ukrainian, Filipino, Indigenous languages in Canada and Indian language politics. No single country dominates the map.
What defines this selection
Archive-scale attention
With 99,235 posts analysed in a single topic, this is a long-running, high-volume observatory of the translation world. The account dates back to 2011, giving the selection the feel of an archive as much as a live feed.
A broad professional lens
The material moves easily between AI translation tools, national dictionaries, literary translation, terminology work, market news and language policy. That range reflects a professional view of language work as both craft and infrastructure.
Mixed-source reliability
The sources span institutions, universities, technology companies, mainstream media and regional outlets. UNESCO, university sites, DeepL, Mistral, CBC, Ouest-France and other sources create a mixed diet of policy, research, product news and local reporting.
Specialist reach, public scale
The audience numbers are substantial, with more than 567,000 views and over 424,000 unique visitors across all topics. The reach suggests sustained interest beyond a narrow specialist circle.
Topics
◍
Metaglossia: The Translation World
This flagship topic covers translation, interpreting, intercultural communication, terminology and lexicography as they unfold in public life. It is especially strong on AI language tools, institutional language policy, minority languages, literary translation and dictionaries.
99.2K posts
en
Deep dives
Where language becomes infrastructure
Charles Tiayon’s main topic reads like a newsroom for the language professions, but with a wider lens than trade news alone. Translation, interpreting, terminology and lexicography appear beside minority-language rights, literary circulation, dictionaries, competitions, and the fast-moving world of AI speech and OCR tools.
Metaglossia: The Translation World
The strongest thread is the collision between language as culture and language as system. A new Maltese digital dictionary, a Doha historical dictionary with text-to-speech, UNESCO and Unicode cooperation, and First Nations language translation all sit in the same frame as DeepL, Mistral OCR, and real-time speech models.
Metaglossia: The Translation World
His editorial voice is not silent aggregation. The analysis points to regular INSIGHT notes, often substantial and quote-based, with an explanatory tone and occasional sharper framing when language politics or AI consequences are at stake. The result is a practical map of a field being reshaped by institutions, software, markets and communities at once."
Courtesy
Scoop.it
06.07.2026
#metaglossia_mundus
#metaglossia
Transient Worlds” because I believe translations by nature are ephemeral, because language is always changing, or vocabulary, or syntax, or diction is always changing. You can have a wonderful translation, but it will only be good for a certain amount of time, and that's not a bad thing. Each generation needs to remake, if the poem is good, their own translation.
"U.S. Poet Laureate who builds bridges across language barriers comes to Oʻahu
Poet Laureate Arthur Sze hikes among ancient indigenous rock carvings that inspired some of his work at La Cieneguilla Petroglyph Site near Santa Fe, New Mexico, March 12, 2026.
Arthur Sze is the 25th U.S. Poet Laureate and a former Merwin Conservancy resident, and he’s in town for two free poetry events this Friday and Saturday.
These events mark the start of his “Words Bridging Worlds” tour, a literary project that offers public events and workshops to share his passion for poetry and translation.
Sze is known for his translations of ancient Tang Chinese poetry into English, and as poet laureate he’s made the theme of sharing and diversifying meaning across languages the focus of his work.
He spoke with HPR to share more about the poetry he’s created and the works that inspire him.
Interview Highlights
On his book, “Transient Worlds”
ARTHUR SZE: In my first year, I assembled this book, "Transient Worlds," on translating poetry, and it features 22 poems from 13 different languages around the world, and it brings poetry from many different languages and cultures into conversation, and I wanted the book to be very inviting, so that instead of most books will say, "Here's the original poem, here's the translation, I did things like, "Here's the original poem in Chinese. Here are three different translations in English. Each one can provide a lens or some vision of what's happening in the original, but no one translation is definitive. There's no such thing as that. And the other thing I wanted to do was make the book very inviting. So in that first zone, I invite the reader to make their own translation of the Chinese poem.
On the arts of translation and poetry
SZE: When I read translations of ancient Chinese poetry translated into English years ago, I felt really uncomfortable because many of the translations had voices. The speakers sounded like they lived 100 years ago, or their use of the language, the diction, felt dated. And I call my book “Transient Worlds” because I believe translations by nature are ephemeral, because language is always changing, or vocabulary, or syntax, or diction is always changing. You can have a wonderful translation, but it will only be good for a certain amount of time, and that's not a bad thing. Each generation needs to remake, if the poem is good, their own translation. So one of the things I'm looking for in a good translation is a voice that is contemporary and alive, and to feel that the language has a kind of rigor and spontaneity behind it. And yes, there's no one way to do it."
Hawaiʻi Public Radio | By Jacob Aloi
Published September 18, 2026
https://www.hawaiipublicradio.org/the-conversation/2026-09-18/u-s-poet-laureate-known-for-building-bridges-across-language-barriers-comes-to-oahu
#metaglossia
#metaglossia_mundus: For content relevance!
Sign Language interpreter for children on Santa Special trains
"Santa trains to get BSL interpreter for first time
The festive trains operate thanks to a team of more than 800 volunteers
A heritage railway will have a British Sign Language interpreter for children on its Santa Special trains for the first time later this year.
The East Lancashire Railway (ELR) said the festive trains operate thanks to a team of more than 800 volunteers to make "a welcoming and magical experience for as many people as possible".
Three years ago it ran its first experience adapted for those with special educational needs, including toned down lighting, quiet carriages and less "hustle and bustle".
ELR chairman, Mike Kelly, said the railway wants "every visitor, whatever their additional needs or challenges" to experience "the wonder and excitement of Christmas".
ELR
Station master Daniel Threadgold, ELR commercial manager Lorna Kelly, BSL interpreter Marie Marshall and Family Care group director Phil Williamson
The interpreter will be sponsored by Chorley-based Family Care, a foster care and special needs provide.
The BSL interpreter will travel on specific Santa Specials run by ELR between Rawtenstall and Bury from 29 November until Christmas Eve.
Interpreter Marie Marshall said she was "really looking forward" to being part of the team.
The ELR was closed to passengers by British Rail in 1972. It was rebuilt later that same decade, continuing into the 1980s, and reopened between Bury and Ramsbottom in 1987. It was later extended to Rawtenstall and Heywood.
The 12.5 mile (20km) heritage route runs through the scenic Irwell Valley, passing through a landscape dotted with industrial relics, factories, fields and cottages."
Emma Stanley
North West
ELR
https://www.bbc.com/news/articles/cje3rwxr5qzpo
#metaglossia
#metaglossia_mundus
The Homeric scholar’s new book on translation is bitingly smart.
"Emily Wilson Chooses Her Words Carefully
A woman with glasses and tattooed arms rests her elbows on the back of a green plush auditorium seat. She leans forward slightly while seated in a large, empty theater filled with rows of matching green velvet chairs...
Is there a scholar better at promoting a book launch than Emily Wilson? Almost a decade ago she revolutionized the promotion of literary translation after the release of her first Odyssey translation. While literary media outlets were breathlessly heralding the arrival of the first-ever translation of the Odyssey into English by a woman, Wilson took to Twitter and showed everyone in granular detail how a translator works, with lengthy threads (still available here, even though Wilson has since deleted her Twitter account) detailing how the choices she’d made differed from earlier renditions.
The threads combined transparency, accessibility, and humor, and they were hugely popular. Whether you came to them as a trained classicist trying to figure out which translation you wanted to teach, or as a newcomer to Homer wondering why another translation was needed at all, Wilson had something to teach you..."
September 18, 2026, 2:00 PM
By Donna Zuckerberg, the author of Not All Dead White Men: Classics and Misogyny in the Digital Age and the forthcoming memoir Antiquated.
https://foreignpolicy.com/2026/09/18/odyssey-emily-wilson-translation-nolan/
#metaglossia
#metaglossia_mundus
Translating Shakespeare Into Material Form
Bespoke textile books translate the structural complexity of the sonnets into tactile, page-based narratives
"Coded Threads: Translating Shakespeare Into Material Form, a Talk by Artist Suzanne Coley
September 23 @ 4:30 pm - 5:30 pm
Free
Suzanne Coley is a book artist, printmaker, and writer whose work engages textile art, printmaking, and embroidered bindings as forms of literary interpretation. Through a signature process she defines as Material Excavation, she recomposes fabric through stitching, layering, and formal transformation, treating material process as a rigorous method of reading Shakespeare’s sonnets. Her bespoke textile books translate the structural complexity of the sonnets into tactile, page-based narratives. Please RSVP to attend the talk.
Coley’s work is held in the collections of Ohio University Libraries, the Smithsonian National Museum of African Art, The National Museum of Women in the Arts, and the Folger Shakespeare Library. Her current project, 52 Weeks of Shakespeare’s Sonnets, explores the sonnets through complex, tactile, page based narratives.
Light refreshments will be served.
Event sponsored by the English Department’s Shakespeare Fund and Ohio University Libraries with support from the Chaddock + Morrow College of Fine Arts and the College of Arts and Sciences."
https://woub.org/event/coded-threads-translating-shakespeare-into-material-form-a-talk-by-artist-suzanne-coley/
#metaglossia
#metaglossia_mundus
Traduire, quelle aventure !
"Atelier traduction de l'italien au français
Le 03/12/2026
Bibliothèque Germaine Tillion | Paris
Gratuit
Traduire, quelle aventure !
Glissez-vous dans la peau d’un traducteur le temps d’un atelier et venez découvrir les joies de la traduction littéraire ! Rendez-vous à la bibliothèque Germaine Tillion le jeudi 3 décembre à 19h.
Accompagnés par Marie Giudicelli, les participants se pencheront sur les pages d’une œuvre littéraire en langue italienne. Pendant deux heures, ils navigueront d’une langue à l’autre à l’aide d’un lexique détaillé, dans un cadre bienveillant et détendu. Cet atelier s’adresse à tous. Aucune connaissance de l’italien n’est requise.
Marie Giudicelli est traductrice littéraire de l’italien et de l’anglais. Elle s’est spécialisée depuis quelques années dans le roman graphique et la bande dessinée. Pour elle qui aime passionnément son métier, les ateliers de traduction sont avant tout des moments de partage, où elle a à cœur de transmettre son amour des langues.
Public adulte. À partir de 15 ans.
Accessible aux personnes à mobilité réduite.
Cet événement a été renseigné par un organisme institutionnel. Date de dernière mise à jour le 20/09/2026."
https://www.jds.fr/paris/activite-experience/ateliers/atelier-traduction-de-l-italien-au-francais-1758720_A
#metaglossia
#metaglossia_mundus
Enregistrement, transcription, compte rendu de réunions, et traduction instantanée : trois fonctionnalités mises en avant
"Nous avons pu découvrir les Shokz OpenFit 2 AI en avant-première à l’IFA 2026. Toujours fidèles au concept d’écouteurs ouverts de la marque, ces nouveaux modèles ajoutent une dimension supplémentaire qui pourrait bien changer leur usage au quotidien.
Présentés pour la première fois hors de Chine à l’IFA 2026, les Shokz OpenFit 2 AI marquent une nouvelle étape pour le spécialiste des écouteurs ouverts. Ils ont non seulement pour particularité de diffuser de la musique tout en laissant l’utilisateur entendre son environnement, mais aussi de se transformer en outil de productivité. Le produit est déjà commercialisé en Chine, mais aucun lancement n’est pour l’heure prévu en Europe.
Ad
Les OpenFit 2 AI embarquent le modèle Qwen et se présentent comme une sorte de boîte à outils IA contrôlable depuis les écouteurs. Les trois fonctionnalités mises en avant par Shokz sont l’enregistrement et la transcription, le compte rendu de réunions et la traduction instantanée.
Jusqu'à 8 h d'enregistrement via les écouteurs ou le boîtier
L’enregistrement peut être déclenché depuis l’un des écouteurs, le boîtier ou l’application Shokz. La marque annonce une portée d’enregistrement allant jusqu’à 5 m dans des environnements contrôlés et jusqu’à 8 h d’enregistrement en stéréo. Un double appui permet également de marquer un passage important pour le retrouver plus facilement par la suite.
Si la captation s’effectue principalement à l’aide des quatre microphones intégrés dans les écouteurs, elle peut aussi être réalisée directement via le boîtier, qui intègre lui aussi un microphone. L’utilisateur peut donc enregistrer lorsque les écouteurs ne sont pas portés, voire lorsqu’ils sont rangés dans leur boîtier. L’enregistrement est ensuite transmis des écouteurs au smartphone, puis directement stocké dans le cloud, avant de pouvoir être consulté au sein de l’app Shokz.
Pour la transcription, Shokz revendique 49 langues et dialectes, répartis en 30 catégories de langues et 19 dialectes chinois. Le français figure bien parmi les langues prises en charge. Le système peut également distinguer plusieurs intervenants afin de faciliter la lecture des transcriptions affichées sur le smartphone.
Après une réunion, l’IA peut générer un résumé avec les principaux sujets abordés, les tâches à effectuer, etc. Shokz annonce que 19 modèles de comptes rendus sont proposés, dont certains adaptés à des domaines comme le droit, la finance, l’informatique, la médecine ou encore le journalisme.
Traduction en face-à-face, directement dans les écouteurs
Les OpenFit 2 AI proposent également plusieurs fonctions de traduction : traduction des appels, traduction en face-à-face et traduction de contenus audio ou audiovisuels. Pour la traduction vocale, Shokz annonce actuellement 11 langues, parmi lesquelles le français, l’anglais, l’allemand, le japonais, le coréen, le russe, le portugais, le thaï, l’indonésien et le vietnamien, en plus du chinois. Les propos de notre interlocuteur, qu’il soit au bout du fil ou directement en face de nous, sont énoncés à voix haute via les écouteurs dans une langue définie au préalable.
Le fait que le français soit déjà pris en charge en Chine ne signifie donc pas que le produit soit immédiatement prêt pour l’Europe. Shokz semble plutôt vouloir valider la qualité et la fiabilité de son expérience multilingue avant de procéder au lancement européen. Pour des fonctions comme la transcription ou la traduction en temps réel, la gestion des accents, des conversations à plusieurs, des contextes professionnels et des différentes variantes linguistiques sont effectivement un enjeu important.
Des OpenFit 2+ boostés à l'IA, sans trop passer à la caisse ?
Outre les fonctions IA, les OpenFit 2 AI reprennent à l’identique la fiche technique des OpenFit 2+. Ils embarquent le principe acoustique DualBoost, qui utilise deux transducteurs par oreillette, tandis que les traitements OpenBass 2.0 et DirectPitch 2.0 renforcent les basses et limitent les fuites sonores. Chaque écouteur pèse environ 9,4 g et se positionne autour de l’oreille à l’aide d’un crochet en titane flexible recouvert de silicone. Surtout, ils demeurent placés à environ un demi-centimètre du conduit auditif, sans y pénétrer, afin d’offrir un confort optimal au quotidien.
Côté autonomie, Shokz annonce jusqu’à 11 h sur une charge et 48 h avec le boîtier. Ce dernier peut être rechargé en USB-C ou via un chargeur sans fil Qi."
Arthur Letang
19/09/26
https://www.lesnumeriques.com/ecouteurs-sans-fil/on-a-vu-les-shokz-openfit-2-ai-ces-ecouteurs-veulent-devenir-votre-traducteur-et-prendre-vos-notes-n262027.html
#metaglossia
#metaglossia_mundus
Appel à la grève des traducteur•trices le 30 septembre 2026
19 septembre 2026UD 93
Partagez cet article
Nous relayons l’appel du collectif En chair et en os, du Stef (syndicat des traducteur·ices de l’édition en France) et du collectif IA-lerte Générale à une grève des traducteurs et traductrices le 30 septembre 2026. Cet appel a été rejoint par Solidaires informatique.
Traducteur·ices : le 30 septembre prochain, on s’arrête et on se mobilise !
Appel du collectif En chair et en os, du Stef, syndicat des traducteur·ices de l’édition en France,
et du collectif IA-lerte Générale
Pour la première fois en 2025, En chair et en os, IA-lerte Générale et l’ATLF organisaient partout en France une grève » des traducteur·ices pour la journée mondiale de la traduction, le 30 septembre. Une grève symbolique – car ce droit n’est toujours pas garanti à notre profession, la plupart d’entre nous travaillant en indépendant·es – mais qui témoignait de notre détermination à défendre nos conditions de travail, à refuser l’IA et son monde et à imposer la reconnaissance de nos métiers.
En 2026, on remet ça : faisons du 30 septembre une journée de lutte collective nationale et internationale, contre l’isolement et le fatalisme. Prenons le temps de nous rencontrer, de nous rassembler, d’échanger, et de nous organiser.
Lire la suite de l’appel :
Appel-a-la-greve-des-traducteur·ices-30-septembre-2026-Télécharger
Le 30 septembre, rendez-vous aux Poussières 1 rue Sadi-Carnot à Aubervilliers : ateliers, table-rondes, cantine, AG, etc. Consulter le programme
Syndicat STEF
Collectif En chair et en os
Appel à la grève de Solidaires informatique
Solidaires Informatique appelle à la grève des travailleur’euses du numérique, des bureaux d’études et du jeu vidéo le 30 septembre 2026, lors de la journée internationale de la traduction, et apporte son soutien à l’appel à la grève et à la mobilisation du collectif En chair et en os, du Stef, syndicat des traducteur·ices de l’édition en France, et du collectif IA-lerte Générale.
Déjà victimes d’une grande précarité, les travailleur’euses du monde de la traduction qui travaillent pour des entreprises de la tech et du conseil subissent doublement l’impact néfaste du recours grandissant aux technologies qualifiées d’intelligence artificielle : augmentation de la cadence de travail, réduction des tarifs et des opportunités d’emploi ou de contrats, transformation forcée de leurs méthodes de travail, utilisation de leur travail pour entraîner des LLM. De plus en plus, les entreprises cessent d’avoir recours à la traduction dite « humaine », pour la remplacer partiellement ou entièrement par de la traduction « par IA ». Ce faisant, les entreprises ne réduisent pas la part humaine : en réalité ils la délocalisent à moindre coût – économique, mais au fort coût écologique, social et humain. Les outils d’intelligence artificielle reposent entièrement sur la surexploitation des travailleureuses du clic des pays du Sud, pour entraîner, valider et corriger leurs modèles. Derrière l’apparente automatisation de ces technologies se cachent des chaînes de sous-traitance complexes, qui permettent aux entreprises de réduire leurs coûts, maximiser leurs profits, tout en maintenant l’opacité de leur chaîne de valeur.
Dans la lignée de notre revendication « Notre syndicalisme face à l’Intelligence Artificielle », Solidaires Informatique soutient l’organisation des travailleur’euses affecté’es face à l’aggravation de leurs conditions de travail dans un contexte de généralisation du recours à l’intelligence artificielle."
https://solidaires93.org/appel-a-la-greve-des-traducteurtrices-le-30-septembre-2026/
#metaglossia
#metaglossia_mundus
Traduction en Français facile à lire et à comprendre (FALC) pour favoriser la compréhension des documents officiels par tous les parents
"Toujours plus de document traduits en Français simple. L’Etat de Fribourg veut rendre accessible des décisions écrites
Une traduction en Français facile à lire et à comprendre (FALC) a été mise en place pour favoriser la compréhension des documents officiels par tous les parents.
Certains documents en provenance de l’État de Fribourg seront désormais traduits en français facile à lire et à comprendre (FALC) au sein des services de l’enseignement obligatoire et spécialisé. L’objectif est de rendre accessible des décisions écrites à tous les parents y compris ceux qui éprouvent des difficultés de compréhensions selon un communiqué de l’État de Fribourg.
Pour se faire, deux collaboratrices spécialisées du Service de l’enseignement obligatoire de langue française (SEnOF) et du service de l’enseignement spécialisé et des mesures d’aide (SESAM) ont suivi une formation par le Bureau du langage simplifié de Pro Infirmis. La démarche s’étend à un nombre toujours plus grand de document avec également une traduction en allemand simple."
https://www.laliberte.ch/articles/regions/canton/l-etat-de-fribourg-veut-rendre-accessible-des-decisions-ecrites-1430211
#metaglossia
#metaglossia_mundus
"L’Institut de traduction de Tunis (ITT) célébrera son 20e anniversaire et la Journée mondiale de la traduction à travers un colloque international consacré au dialogue interculturel, les 30 septembre et 1er octobre 2026 à la Cité de la Culture.
Intitulé « Esthétique et éthique de la traduction dans le contexte du dialogue interculturel », l’événement réunira des chercheurs, universitaires et experts de plusieurs pays, notamment de Tunisie, d’Algérie, d’Italie, de Russie et du Portugal.
Le programme prévoit quatre sessions scientifiques et un forum thématique portant notamment sur l’éthique textuelle, la traduction poétique, l’éco-traductologie ainsi que les enjeux liés à l’intelligence artificielle et à la traduction automatique.
Fondé en 2006, l’ITT entend également dresser, à travers cette rencontre, le bilan de deux décennies d’activités consacrées aux échanges linguistiques et culturels."
2026/09/18 23:25
https://www.mosaiquefm.net/amp/fr/national-tunisie/1530712/l-institut-de-traduction-de-tunis-celebre-ses-20-ans
#metaglossia
#metaglossia_mundus
"Visio-interprétation ou langue des signes
Ce décret entrera en vigueur « le premier jour du sixième mois suivant sa date de publication », donc le 1er mars 2027. Il précise que pour appliquer ces dispositions, les personnes sourdes et malentendantes doivent pouvoir disposer « d’un dispositif d’accueil adapté permettant la traduction simultanée écrite et visuelle », dispositif qui peut prendre plusieurs formes : « la transcription écrite instantanée » , le recours à un interprète en langue des signes (en présentiel ou en visio), l’échange avec un agent maîtrisant la langue des signes ou… « tout autre moyen permettant à l’usager d’accéder à des fonctionnalités équivalentes ».
Le décret précise que si un tel dispositif n’est pas « immédiatement disponible », la personne chargée du service public « organise, dans un délai raisonnable, les modalités permettant un échange dans des conditions adaptées ».
3 000 euros par an, a minima
Comme toutes les mesures améliorant l’accès au service public des personnes en situation de handicap, il s’agit évidemment d’une mesure positive. Mais comment la mettre en œuvre ? Et à quel prix ?
Ce sont les questions qu’ont posées les représentants de l’AMF, lors de l’examen de ce projet de décret par le Conseil national d’évaluation des normes, en juillet dernier. À cette occasion, le ministère concerné a indiqué avoir procédé au début de l’année à une consultation auprès des services publics de l’État qui ont déjà mis en place un tel service, qui a permis de mettre en lumière que la majeure partie d’entre eux ont recours à des services dits de « visio-interprétation ». Il s’agit de mettre en contact par visioconférence la personne sourde ou malentendante et un interprète s’exprimant en langue des signes. Un bon nombre d’entreprises proposent de tels services.
Mais ils ont un coût élevé, que la consultation organisée par le ministère a permis d’estimer « entre 20 000 et 80 000 euros HT par an », selon les forfaits. Le gouvernement estime néanmoins – sans que l’AMF ait bien compris sur quoi repose cette estimation – que pour les communes, un montant moyen de 3 000 euros est à envisager.
L’AMF, qui n’a pas rejeté par principe ce décret, dans la mesure où il va dans le sens d’une plus grande accessibilité des services publics, s’est en revanche interrogée sur ce coût estimé de 3 000 euros, qui serait démultiplié en fonction du nombre de points d’accueil dans les communes – celles-ci, rappelant l’association, étant « responsables d’une grande majorité des services publics qui accueillent des usagers ».
Et si les communes font le choix de l’accueil par des agents maîtrisant la langue des signes, cela engendrera là encore des coûts de formation non négligeables.
La question reste donc entière de savoir comment les communes et EPCI vont pouvoir financer ce nouveau service, alors qu’aucun fonds de soutien en la matière n’est prévu par le gouvernement, et au moment où les injonctions à réduire les dépenses des collectivités se font plus pressantes que jamais.
On ignore également, à ce jour, quel accompagnement sera mis en place par l’État, d’un point de vue opérationnel, pour aider les communes – notamment les plus petites – à mettre en place ce service dans un délai si restreint. Ces dispositions législatives ayant été adoptées il y a trois ans, il est bien dommageable que l’anticipation, une fois de plus, n’ait pas été au rendez-vous, avec la parution tardive d’un décret qui ne laisse que six mois aux communes pour s’organiser.
Suivez Maire info sur Twitter : @Maireinfo
https://www.maire-info.com/accessibilita%C2%A9/dans-six-mois-toutes-les-communes-devront-a%C2%AAtre-a%C2%A9quipa%C2%A9es-pour-accueillir-les-usagers-sourds-ou-malentendants-article-31089
#metaglossia
#metaglossia_mundus
DeepL Voice améliore la communication multilingue en temps réel
"DeepL, leader en intelligence artificielle linguistique, a annoncé des mises à jour majeures de sa technologie d'IA vocale. Ces nouvelles fonctionnalités visent à préserver les qualités uniques de la voix de chaque locuteur lors des traductions en temps réel. Cette innovation permet des conversations plus naturelles entre les langues, en améliorant la restitution du ton, des émotions et des expressions.
L'entreprise a également lancé une nouvelle application de bureau permettant des réunions multilingues fluides grâce à la traduction vocale en temps réel, désormais disponible dans plus de 30 langues. Cette avancée est particulièrement avantageuse pour les utilisateurs de Zoom, Microsoft Teams et Google Meet, car elle favorise des interactions interlinguistiques naturelles.
Des entreprises de divers secteurs intègrent DeepL Voice pour améliorer la collaboration internationale et conquérir de nouveaux marchés. DeepL assure notamment les traductions en temps réel lors de Salesforce Dreamforce 2026, un événement d'envergure proposant de nombreuses sessions.
R. H."
https://www.webdisclosure.com/article/deepl-etr-deepl-voice-ameliore-la-communication-multilingue-en-temps-reel-X33IYU1kBb4
#metaglossia
#metaglossia_mundus
La première traduction intégrale du Code Noir en créole haïtien. 341 ans après Versailles.
"James Fleurissaint : 341 ans de silence brisé, le Code Noir parle enfin créole
Le 28 octobre paraîtra un événement que j'attends depuis des années : la première traduction intégrale du Code Noir en créole haïtien. 341 ans après Versailles.
Ce n'est pas une traduction. C'est une victoire.
1. Mémoire
Le Code Noir n'a pas seulement volé le travail de nos ancêtres. Il a volé leur mémoire.
Mars 1685. 60 articles. L'article 44 dit : l'esclave est un bien meuble. On le vend comme une chaise. L'article 2 dit : il doit être baptisé et instruit dans la religion catholique. Traduction : on efface son nom, sa langue, ses dieux.
Pendant 341 ans, ce texte est resté en français ancien, illisible pour les descendants de ceux qu'il a transformés en objets. On nous a enseigné 1492, on ne nous a jamais enseigné 1685. Résultat : nous avons un cimetière à l'intérieur de nous. Une mémoire bledigo, bleue, enflée de coups qu'on ne sait pas nommer.
Traduire le Code Noir en créole, c'est rendre son nom à la blessure.
2. Langue
Le créole est né dans l'habitation que le Code Noir organisait. C'est la langue inventée par les esclaves pour se comprendre entre eux quand le maître interdisait leurs langues africaines.
Et pourtant, cette langue n'a jamais eu force de loi.
La Constitution haïtienne de 1987 dit article 5 : créole et français sont langues officielles. Article 40 : toutes les lois doivent être publiées dans les deux langues. Dans les faits, les lois sont en français, le tribunal parle français, l'école punit encore l'enfant qui parle créole.
Le français a toujours eu le monopole de la loi. Le créole a eu le monopole de la vie.
J'ai traduit les 60 articles mot à mot depuis l'édition originale de Versailles, en créole haïtien standard, sans y laisser un seul mot de français. Pour prouver qu'on peut dire le droit le plus violent dans notre langue maternelle.
3. Victoire
Le 28 octobre, pour la première fois, la loi qui nous a niés va parler notre langue.
Ce jour-là, le créole ne sera plus seulement la langue de la cuisine, de la blague, de la chanson. Il deviendra la langue qui juge l'histoire. La langue qui dit : voilà ce qu'on nous a fait, avec les mots exacts de ceux à qui on l'a fait.
Le français avait servi à donner force de loi à la violence. Le créole va maintenant lui donner force de mémoire.
C'est ça la victoire. Ce n'est pas moi qui gagne. C'est la langue qui revient chercher ses morts.
La traduction intégrale paraîtra le 28 octobre en format papier, dans une grande maison d'édition présente en France, au Canada et en Haïti.
Godson Moulite"
https://blogs.mediapart.fr/jamesfleurissaint1804/blog/170926/james-fleurissaint-341-ans-de-silence-brise-le-code-noir-parle-enfin-creole
#metaglossia
#metaglossia_mundus
"Je traduis, donc nous sommes : Détours de la traduction (en ligne)
Le 2 Octobre 2026, 4 Décembre 2026, 5 Février 2027, 5 Mars 2027, 2 Avril 2027, et le 14 Mai 2027
Lien : https://univ-reunion-fr.zoom.us/j/86527546059
Publié le 17 Septembre 2026 par Léo Mesguich (Source : Bénédicte Letellier)
Webinaire de recherche, 2026-2027
Organisé par Laëtitia Saint-Loubert (CRINI, Nantes Université) et Bénédicte Letellier (DIRE, Université de La Réunion)
En ligne, d’octobre 2026 à mai 2027
Présentation
Et si la ligne droite était l’un des grands mythes de la traduction ?
Un texte ne va presque jamais simplement d’une langue à une autre. Il bifurque, transite, revient sur ses pas, change de mains, de voix, de support ou de statut. Il passe par une langue-relais, une retraduction, une adaptation, une institution, un éditeur, une communauté de lecteurs, parfois un dispositif algorithmique. Il peut être déplacé, censuré, réapproprié, réécrit — et c’est souvent dans ces écarts que se joue sa puissance de transformation.
Le séminaire propose cette année de placer le détour au centre de nos réflexions. Il s'agira de considérer le détour non seulement comme déviation accidentelle par rapport à une traduction supposée directe, mais aussi comme forme constitutive de l’acte traductif. Plus largement que la seule indirect translation, le détour désigne ici l’ensemble des médiations, déplacements et négociations qui s’interposent entre un texte et ses devenirs. Il invite ainsi à déplacer quelques-unes des catégories les plus familières des Translation Studies — original et dérivé, auteur et traducteur, fidélité et trahison, centre et périphérie — pour suivre ce qui arrive aux textes lorsqu’ils circulent réellement.
Mais le détour ne fait pas que transformer les textes : il transforme les relations. C’est le pari contenu dans la formule qui donne son titre à cette deuxième année : « Je traduis, donc nous sommes ». Le « nous » de la traduction n’existe pas nécessairement avant elle. Il se constitue dans l’acte même qui met en relation des langues, des temporalités, des expériences et des mondes sans les réduire à l’identique. Traduire ne serait donc pas simplement franchir une distance entre des communautés préexistantes, mais participer à la fabrication toujours instable de leurs relations. C’est précisément l’une des hypothèses centrales du séminaire : le « nous » n’est pas le préalable de la traduction, mais peut en être l’un des effets.
Cette hypothèse prend une acuité particulière au moment où les conditions mêmes de la médiation se transforment. Traductions collectives et communautaires, circulations transnationales, rétrotraductions, pratiques plurilingues, plateformes numériques et intelligence artificielle redistribuent aujourd’hui les places de l’auteur, du traducteur, du texte et de ses publics. Le détour permet alors de poser une question commune à des objets très différents : qui — ou quoi — traduit lorsque les textes circulent, et quel monde commun ces opérations contribuent-elles à produire ?
À partir d’études consacrées à l’arabe soudanais, aux communautés de rétrotraduction, aux trajectoires de traducteurs, aux circuits de la littérature traduite en Irlande, aux médiations linguistiques dans les contextes policiers français ou encore aux circulations franco-japonaises de la poésie, le séminaire explorera ainsi le détour comme opérateur critique pour penser ensemble traduction, médiation, relation et circulation.
Cette nouvelle saison prolonge une première année du séminaire, « Traduction et détournements : une culture universelle du piratage ? », qui a accueilli en 2025-2026 Lily Robert-Foley, Claudine Le Blanc, Michael Cronin, Coraline Jortay, Samah Selim, Cécile Brochard et Joaquin Ruiz Zubizarreta.
Programme 2026-2027
Vendredi 2 octobre 2026
Xavier Luffin (Université libre de Bruxelles)
« Traduire l’arabe, l’arabe soudanais, la culture soudanaise, ou les trois »
Vendredi 4 décembre 2026
Mathias Verger (Université Paris 8)
« Le nous et le noûs de la rétrotraduction : des communautés traductives singulières et plurielles »
Vendredi 5 février 2027
Nathan Dize (Washington University in St. Louis)
« Learning Backwards: A Translator’s Path Back to the Beginning »
Vendredi 5 mars 2027
Tim Groenland et Michaela Králová (University College Dublin)
« Contemporary Translated Literature in Ireland: Mediators, Circuits, and Detours »
Vendredi 2 avril 2027
Audrey Cartron (Nantes Université)
« Exposure to English in French Police Contexts: Linguistic, Discursive and Cultural Detours and Mediations »
Vendredi 14 mai 2027
Oriane Chevalier (Université Clermont Auvergne)
« Nous traduisons, donc nous sommes : la poésie japonaise entre les doigts de Judith Gautier et de Saionji Kinmochi »
Organisation
Laëtitia Saint-Loubert
CRINI – Nantes Université
Bénédicte Letellier
DIRE – Université de La Réunion
Le séminaire se tient entièrement en ligne sur Zoom.
Horaire : premier vendredi de chaque mois à 14h (Heure de Paris)
Lien de connexion : https://univ-reunion-fr.zoom.us/j/86527546059
Contact :
Laetitia.Saint-Loubert@univ-nantes.fr
benedicte.letellier@univ-reunion.fr
Responsable :
Bénédicte Letellier et Laëtitia Saint-Loubert"
https://www.fabula.org/actualites/136504/je-traduis-donc-nous-sommes-detours-de-la-traduction.html
#metaglossia
#metaglossia_mundus
"Court interpreting may look like a simple exchange of words, but what happens when there is no direct translation, no advance access to the conversation, and no room for misunderstanding?
Today, host Mary Anne Sioco continues her conversation with English–Chinese court interpreter Angel Kao. This time, they look beyond language ability to explore the concentration, preparation, and quick thinking the profession demands. Angel takes us behind the scenes of a courtroom assignment, where an interpreter must listen closely, preserve the speaker’s true meaning, navigate unfamiliar legal terms, and remain aware of far more than the words being spoken.
But the challenges do not always end when the hearing is over. Angel also opens up about a lesser-known side of the profession—one that raises important questions about personal safety, boundaries, and whether interpreters receive enough recognition and protection for the responsibilities they carry.
If you missed Part One, go back and discover the unexpected personal journey that led Angel from studying English in Taiwan and living overseas to finding her place inside Taiwan’s courtrooms.
Then join us for Part Two, as we uncover what it means to speak for someone when every word and every decision matters."
Radio Taiwan International (RTI) is Taiwan`s national broadcaster that provides engaging news, audio and video features about Taiwan. RTI is a unique window into the heart and soul of Taiwan society.
https://www.rti.org.tw/en/programnews?uid=4&pid=113912
#metaglossia
#metaglossia_mundus
Radio Taiwan International (RTI) is Taiwan`s national broadcaster that provides engaging news, audio and video features about Taiwan. RTI is a unique window into the heart and soul of Taiwan society
Firm is fired for poor translation on aid forms for Alaska Natives
Berkeley company apologizes for errors in FEMA documents for typhoon damage
"ANCHORAGE — After tidal surges and high winds from the remnants of a rare typhoon caused extensive damage to homes along Alaska’s western coast in September, the U.S. government stepped in to help residents — largely Alaska Natives — repair property damage.
Residents who opened Federal Emergency Management Agency paperwork expected to find instructions in Alaska Native languages such as Yup’ik or Inupiaq on how to file for aid. Instead, they were confronted with a number of bizarre phrases.
“Tomorrow he will go hunting very early, and will [bring] nothing,” one passage read. The translator randomly added the word “Alaska” in the middle of the sentence.
“Your husband is a polar bear, skinny,” another said.
Yet another was written entirely in Inuktitut, an Indigenous language spoken in northern Canada, far from Alaska.
FEMA fired the California company hired to translate the documents once the errors became known, but the incident was an ugly reminder for Alaska Natives of the suppression of their culture and languages from decades past.
FEMA immediately took responsibility for the translation errors and corrected them, and the agency is working to make sure it doesn’t happen again, spokesperson Jaclyn Rothenberg said. No one was denied aid because of the errors, which were first reported by KYUK Public Media in Bethel, Alaska.
Caroline Lee, chief executive of Accent on Languages, the Berkeley company that produced the mistranslated documents, said in a statement: “We make no excuses for erroneous translations, and we deeply regret any inconvenience this has caused to the local community.”
She said the company would refund FEMA the $5,116 it received for the work and conduct an internal review to ensure it never happens again.
Lee did not respond to follow-up questions, including how the mistaken translations occurred.
For Tara Sweeney, an Inupiaq who served as an assistant secretary of Indian Affairs in the U.S. Interior Department during the Trump administration, the incident was another painful reminder of steps taken to prevent Alaska Native children from speaking Indigenous languages.
“When my mother was beaten for speaking her language in school, like so many hundreds, thousands of Alaska Natives, to then have the federal government distributing literature representing that it is an Alaska Native language, I can’t even describe the emotion behind that sort of symbolism,” Sweeney said.
She called for a congressional oversight hearing to uncover how long and widespread the practice of outsourcing translations has been used throughout government.
“These government contracting translators have certainly taken advantage of the system, and they have had a profound impact, in my opinion, on vulnerable communities,” said Sween- ey, whose great-grandfather, Roy Ahmaogak, invented the Inupiaq alphabet more than halfa century ago.
She said his intention was to create the characters so “our people would learn to read and write to transition from an oral history to a more tangible written history.”
U.S. Rep. Mary Peltola (D-Alaska), who is Yup’ik and last year became the first Alaska Native elected to Congress, said it was disappointing that FEMA missed the mark with these translations but didn’t call for hearings.
“I am confident FEMA will continue to make the necessary changes to be ready the next time they are called to serve our citizens,” she said.
About 1,300 people have been approved for FEMA assistance after the remnants of Typhoon Merbok created havoc as it traveled about 1,000 miles north through the Bering Strait, potentially affecting 21,000 residents. FEMA has paid out about $6.5 million, Rothenberg said.
Preliminary estimates put overall damage at more than $28 million, but the total is likely to rise after further assessment after the spring thaw, said Jeremy Zidek, a spokesperson for the Alaska Division of Homeland Security and Emergency Management.
The poorly translated documents, which did not create delays or problems, were a small part of efforts to help people register for FEMA assistance in person, online and by phone, Zidek said.
Central Alaskan Yup’ik is the largest of the Alaska Native languages, with about 10,000 speakers in 68 villages across southwest Alaska. Children learn Yup’ik as their first language in 17 of those villages. There are about 3,000 Inupiaq speakers across northern Alaska, according to the language center.
Thiessen writes for the Associated Press."
https://enewspaper.latimes.com/infinity/article_share.aspx?guid=bb9f6732-d431-4302-845e-a9bb78d93d82
#metaglossia #metaglossia_mundus
"A plan by the Literature Translation Institute of Korea (LTI Korea) to establish an independent graduate school dedicated to Korean literary translation in Seoul has been delayed as it went against the government's initiative for "balanced regional development." The institution is now reviewing an interim master's program with a national university outside Seoul while continuing to pursue the longer-term goal of a standalone facility.
The state-run institute under the Ministry of Culture, Sports and Tourism has long operated a translation academy that has nurtured translators who bring Korean literature to wider international audiences. Its key limitation, however, has been that the program falls outside the official degree education system. To address that, the institute planned to open an independent graduate school in Seoul in September 2027.
Those plans have since been revised, according to government officials and other sources, and it is now in talks with regional universities to launch a master's degree program as an interim step instead.
“As a transitional measure, we have revised our plans and are now seeking to start operating this master’s program in collaboration with a regional national university in September next year,” an LTI Korea official told The Korea Times. “As for establishing an independent school, we have not abandoned the plan despite the current circumstances. We intend to continue pursuing it after putting better conditions in place for what’s required for establishment.”
No specific partner school has been determined yet, she added.
The original plan called for remodeling LTI Korea’s headquarters in southern Seoul to serve as the school’s educational facility. It was revised after two months of intensifying discussions among representatives of LTI Korea and the culture and education ministries, she said.
The change came amid the Lee Jae Myung administration’s emphasis on balanced regional development, under which the capital region is seen as already concentrating too much of Korea’s population and resources. The Ministry of Education’s university establishment and restructuring review committee recently expressed the view that the new graduate school should not be established in Seoul.
“My understanding is that members of the committee expressed the view that, given the current strong emphasis on balanced regional development, it would be more desirable to collaborate with institutions in less central regions for the program rather than establish it in Seoul,” an education ministry official said.
LTI Korea began forming a committee in April to advance the translation school plan under the revised Literature Promotion Act, which came into force last year. The envisioned school would have expanded its existing Translation Academy into an accredited master’s degree program dedicated to translating Korean literature.
Unlike other graduate schools of translation and interpretation, the proposed institution would focus specifically on literary translation. The initial plan called for an annual intake of 60 students, including 45 international students.
The project has been viewed as part of a broader effort to cultivate professional translators as Korean literature gains a wider international readership, particularly following novelist Han Kang’s 2024 Nobel Prize in literature.
Many experts in the industry see Seoul as the preferred location for the proposed graduate school, citing concerns that a regional location could make it harder to recruit both students and leading instructors from overseas.
“If the school is located outside Seoul, there would be limits in attracting talented foreign students,” one expert said. “There would also be constraints in inviting teachers, who would have to travel to a regional campus.”
Kwak Hyo-hwan, a former LTI Korea president who led efforts to establish the school, expressed disappointment over the revised plan.
“It is regrettable. If the government’s regional relocation policy takes priority over everything else, there is little more to say except that it is regrettable,” he said.
https://www.koreatimes.co.kr/lifestyle/books/20260914/plan-for-literary-translation-graduate-school-delayed-over-regional-development-policy
#metaglossia
#metaglossia_mundus
A Southeast European studies seminar with translator Ellen Elias-Bursać
"Translating War: Damir Karakaš' Celebration and Blue Moon
16 Oct 2026, 18:00 – 20:00
Masaryk room
14-16 Taviton street
London
WC1H 0BW
United Kingdom
Book now
Capturing the invisible hand of yesterday’s fascism for today’s readers
“Celebration”
In four fragmentary chapters set in the mountainous Lika region of Croatia, Mijo’s life unfurls not with the chronology of time but in episodes, each in flow with the textural grain of this remote earth. With sparest of prose, Karakaš has crafted an exquisite sense of place. But with that comes terrifying violence and the evolution of a fascist. As an unwilling combatant on the run in “The House”, Mijo has not strayed far from home. From his hiding place in the woods above their cabin he observes through the cross hairs of his rifle his beloved wife go about her daily routine. Autumn is advancing and Drenka has agreed that at night it might be safe for his return to the crawl space beneath the house. In “Dogs” Mijo is a child working on his family’s land the day the local authorities pass an edict that all dogs in the valley must be killed. Around the table, his father and mother evaluate a warm life in terms of raw pragmatism, and the journey Mijo decides to make will become his pilgrimage to adulthood. Mijo arrives at the door of his girlfriend, Drenka, on a day of festivity in the third, eponymous chapter, “Celebration”. The couple plans to walk through summer fields to the revelry in town. With her brother as chaperone, the two lovers lag behind. The contours of the horizon have shifted with the seasons and their direction becomes as uncertain as the events that will unfurl at the festivity awaiting them. “Father” concludes the work with devastating finality.
“Blue Moon”
Meet Čarli, a young rockabilly fanatic at the end of the Eighties. A failed student, he cares deeply about his pompadour hairdo, symbolic of his desire for an identity shift away from his peasant roots, and he wears it with obsessive, vain pride. The quiff is his crown, his best self. Čarli searches for mirrors in the street in order to admire the quiff at all times, has located shop windows that reflect the quiff at particularly good angles, spends much time grooming it, and has even devised a way to sleep (with three carefully positioned pillows) to minimize quiff “ruffle”. And it is this very hairstyle that his father, a man so violent that he borders on psychopathic, hates with a vengeance. Following a bloody fist fight between father and son, it is clear that the claustrophobic and violent darkness that looms over Čarli’s childhood home, and Yugoslavia at large, will eventually devour him and those of his generation. BLUE MOON delves into the inability of a world that is approaching a new war, to self-reflect. Through Čarli’s unlikely relationship with Eli, a capable city girl and Leonard Cohen fan, the arteries of the novel are themes of masculinity and violence, and the traditional, conservative preference, in the face of difference, towards selfpreservation. A doubling down into homogeneity. Karakaš is a master of escalating conflict. He draws a stifling darkness around the edges of BLUE MOON building mounting pressure in ways simultaneously surreal, humorous, and suffocating. And it is this dark comedy which is the novel’s extraordinary watermark, a first indicator of his writing to come.
This event will be chaired by Prof. Eric Gordy, UCL SSEES.
Translator
Ellen Elias-Bursać is an award-winning translator of Bosnian, Croatian, and Serbian, who has served as the president of the American Literary Translators’ Association. Between 1972 and 1990 she lived in Zagreb and for six years she worked in the English Translation Unit of the International Criminal Tribunal for the former Yugoslavia in The Hague, Netherlands. Her recent translation from the Croatian into English of Celebration by Damir Karakaš was shortlisted for the EBRD Literature Prize and was a Best Book of 2024 for The Paris Review. Her translation last year of Karakaš’ widely popular second novel, Blue Moon, won an English PEN Award.
Author
Damir Karakaš is a Croatian author, playwright, musician and journalist, winner of the prestigious Meša Selimović award, among many others. Born in 1967 in the remote, mountainous region of Lika, he later studied law and agronomy in Zagreb. During the 1990s he worked as a war reporter from the front lines in Croatia, Bosnia and Kosovo for national daily newspaper Večernji list. His novels have been translated into French, German, Italian, Czech, Slovak, Macedonian, Slovenian, and Arabic. His plays have been performed across Croatia, Serbia, Germany, Chile and Ivory Coast.
Further information
Ticketing
Pre-booking essential
Cost
Free
https://www.ucl.ac.uk/arts-humanities/events/2026/oct/translating-war-damir-karakas-celebration-and-blue-moon
#metaglossia
#metaglossia_mundus
"Analyst(s): Keith Kirkpatrick
Publication Date: September 14, 2026
Zendesk is bringing AI-powered, two-way translation to live contact center calls, letting customers and agents speak in their own languages without interpreters or dedicated language queues. The move goes after a real staffing bottleneck, but the real test is whether machine translation holds up on the calls where getting it right actually matters.
What Is Covered in This Article:
Zendesk’s new Real-Time Voice Translation capability for Zendesk Contact Center, including how the translation and admin controls work.
The staffing and routing problem the feature is built to solve, and the scale Zendesk attaches to it.
How the announcement lines up against Genesys’ AI Control Plane news and Verint’s latest data on AI trust in the contact center.
What still needs to be proven before real-time voice translation earns a place on the calls that matter most.
The News: Zendesk introduced Real-Time Voice Translation for Zendesk Contact Center on September 10, 2026, adding AI-powered, two-way translation to live calls so customers and agents can speak in the language that feels natural to them. Zendesk’s rationale centers on scale: voice still accounts for 40% of contact center volume, according to Zendesk research, and language gaps make those calls harder to staff and route. The company’s pitch is that businesses can offer voice support in more languages without hiring interpreters, building language-specific queues, or pushing customers with complex or urgent issues toward slower channels such as email or SMS.
Mechanically, agents can turn translation on or off mid-call, and administrators control which languages are available and whether the system retains the original audio, the translated audio, or both. That detail matters for compliance review as much as for quality control. The capability is built into existing Contact Center workflows rather than bolted on as a separate tool, so agents keep working the same queues and scripts. Per CX Today’s initial reporting, the feature launches supporting 13 languages: Chinese, English, French, German, Hindi, Indonesian, Italian, Japanese, Korean, Portuguese, Russian, Spanish, and Vietnamese. Real-Time Voice Translation becomes available in October 2026 through a closed Early Access Program for eligible Zendesk Contact Center Native customers.
Shashi Upadhyay, Zendesk’s President of Product, Engineering, and AI, framed the release around exactly this kind of call: “Customers should not have to choose between speaking the language that feels most natural to them and getting the support they need. When an issue is complex, sensitive, or urgent, people want to be understood, not routed somewhere else. Real-Time Voice Translation helps businesses make that kind of voice support available to more customers, while agents stay focused on solving the problem.”
Will Real-Time Voice Translation Solve the Contact Center’s Language Problem?
Analyst Take: Zendesk is going after a cost line every contact center leader can recite from memory: keeping enough multilingual agents on staff to match demand that shifts by region, time zone, and season. That is a staffing and economics problem before it is an AI problem, and it is the right one to target. Hiring a Portuguese-speaking agent to cover a queue that gets three calls a week is not a viable business decision, and the alternative — a queue transfer, a hold, or a shift to email for a customer who called because their issue was urgent — is a worse experience than an imperfect but immediate translation.
A Sensible Problem to Solve, With an Unproven Answer
The open question is quality. Real-time speech translation has to convey tone, urgency, and technical detail across languages in a live conversation, with no chance to pause and rephrase as a written chatbot exchange allows. Zendesk’s own framing (that this matters most when an issue is “complex, sensitive, or urgent”) describes precisely the calls where a mistranslated word carries the highest cost: a billing dispute, a safety complaint, a cancellation threat. Zendesk has not published accuracy data against these kinds of calls, and the feature’s value is unproven until it does.
Retention Controls Are Smart; Compliance is the Harder Question
Letting administrators choose whether to keep original audio, translated audio, or both is the correct design decision, and it signals that Zendesk built this with enterprise compliance teams in mind rather than as an afterthought. That same control previews a harder conversation still ahead. Call recording consent rules vary by state and country, and a translated recording raises questions about whose voice is recorded, what a regulator or a court treats as authoritative, and how long either version can be retained. Those questions fall to Zendesk’s customers, since compliance obligations rest with whoever deploys the feature in a given jurisdiction.
Language Coverage isn’t the Competitive Battlefield; Trust Is
Zendesk’s announcement landed the same week Genesys unveiled an AI Control Plane along with Navigator and Orchestrator tools aimed at coordinating AI agents, human employees, and workflow across the contact center. That is a bet on orchestration rather than on language coverage, and the contrast is the more useful lens on where this market is actually competing.
Verint’s State of Contact Center AI 2026 report found that 87% of contact center leaders plan to increase spending on automated agent assistance, and 94% of organizations believe AI has improved agent performance. Only 62% of consumers say they have noticed a positive impact. That 32-point gap is the number vendors should be working toward closing, and it will determine whether real-time voice translation is adopted as a trusted channel or quietly routed around by customers who would rather wait for a human who speaks their language.
What to Watch:
Whether Zendesk publishes accuracy or quality benchmarks against human interpreters, particularly on nuance, idiom, and emotionally charged conversations, once Early Access customers can speak to outcomes.
Whether the language list Zendesk ships with covers the routes actually driving the multilingual staffing problem for large, distributed contact centers, or leaves gaps in high-volume languages.
Whether Genesys, NICE, and Five9 respond with real-tKeith Kirkpatrick, VP & Research Director, Enterprise Software & Digital Workflows at Futurum, shares his insights on Zendesk’s new real-time voice translation feature for contact centers and what it will take to earn customer trust.ime translation capabilities of their own, rather than continuing to concentrate on orchestration and agent-assist tooling.
Whether enterprise buyers in regulated industries push back on the audio-retention options once legal and compliance teams examine consent rules and cross-border data handling.
You can read the full release at Zendesk’s website."
https://futurumgroup.com/insights/will-real-time-voice-translation-solve-the-contact-centers-language-problem/
#metaglossia
#metaglossia_mundus
"English is taking up an increasingly prominent place in Algeria’s education system at the expense of French – a shift that has stirred controversy across the North African country.
This school year, three years after its introduction in primary school, English is becoming the first foreign language taught to pupils, starting from the third grade.
“We have decided to teach English in the third year of primary school, and French (alongside English) starting in the fourth year. In third year, only English will be taught,” announced the Algerian education minister in July.
Until 2023, French was the only foreign language taught in primary school, starting in third grade. This had been the case since Algeria’s independence in 1962. But following a decision by President Abdelmadjid Tebboune, English – previously introduced only at secondary school level – was added, starting in the third year of primary school, alongside French.
Tebboune argued that English is the language of science and it is therefore important to teach it in school.
The place of English will also be strengthened in secondary school through an increase in teaching hours and a higher coefficient in the end‑of‑cycle exam.
In high school, French teaching has not yet been modified. But English will be further developed, the education minister said.
Although the government’s announcements are significant, teachers’ associations have not commented. The main teachers’ union, Cnapeste, was dissolved last July by the authorities for violating regulations on the right to strike. The union rejects the accusations, viewing the procedure as an attempt to neutralise an organisation known for its mobilisations.
In the past, Cnapeste had called the promotion of English teaching a hasty decision, arguing that teachers are not sufficiently trained to handle the subject.
On the contrary, the National Federation of Parents of Students Associations – which is generally aligned with the ministry’s position – expressed satisfaction with the reform, considering English “the language that gives access to science and technology, and therefore, ultimately, to better study and job prospects for pupils”.
‘How will the state manage to provide every school with qualified English teachers in sufficient numbers?’
– Rachida Mahi, Algerian parent
Not all parents, however, agree. Rachida Mahi, a 36‑year‑old engineer living in Algiers, regrets a rushed decision that risks having a disastrous impact on children’s schooling.
She worries about the future of her son, who is starting third grade and will begin learning English this school year.
“How will the state manage to provide every school with qualified English teachers in sufficient numbers? Where does it plan to find them?” she told Middle East Eye.
Like many Algerians of her generation, Mahi speaks very little English.
“It was a secondary language in our schooling, unlike French. I did my engineering studies in French,” she said, adding that she is panicked at the idea of not being able to help her son with his English homework.
‘Improvised measures’
At the end of French colonisation in 1962, Arabic was established as Algeria’s national and official language. To definitively erase the colonial linguistic legacy, education was gradually Arabised. This policy reached its peak in 1991 with the adoption of a law generalising Arabic in all areas of public life and education.
But Arabisation was not entirely successful.
Why does Arabic have such an image problem in France?
Read More »
At university, only law and social sciences were Arabised. Scientific fields such as medicine or engineering continued to be taught in French, due to the lack of suitable teaching resources and the shortage of teachers trained in Arabic. Further, leading scholarly journals are still largely accessible only in foreign languages – primarily French and English.
President Tebboune chose in 2025 to progressively replace French with English in medical studies, calling it a “wise” decision that would align Algerian higher education with international standards and global scientific research, where English dominates.
Salah Derradji, a specialist in language didactics and former university rector, is sceptical about the new measure. He told MEE that he has long supported the generalisation of English at university and its gradual introduction at other levels of education. But, he warned, “This project must be carried out thoughtfully. And that is not the case today.”
He argues that the Algerian president made “a political decision” and believes that it is impossible to apply such a measure overnight “with a magic wand”.
“You cannot train teachers in a few weeks and send them into classrooms. You first need qualified trainers,” he said, regretting the absence of broad consultation involving specialists.
Sharing his view, Rabeh Sebaa, former professor of sociology and anthropology at the University of Oran, tells MEE that “Algeria does not have the means to teach English – or teach in English – whether in schools or universities.”
In 2022, as it prepared to introduce English in primary school, the Algerian education ministry announced a major recruitment campaign for 60,000 teachers, targeting university graduates in English language and literature, as well as interpretation.
‘This project must be carried out thoughtfully. And that is not the case today’
– Salah Derradji, language didactics specialist
The operation, however, drew criticism from education unions, who denounced the recruitment of people without pedagogical experience and who were trained hastily.
“My daughter, who is entering secondary school this year, still does not express herself well in English,” Hamid Hani, a 42‑year‑old parent from Bejaia, told MEE. “She had the same English teacher for two years – a young woman, inexperienced and unable to manage a classroom.
“This teacher admitted to me during a meeting that she had to manage on her own to find teaching materials and organise her lessons.”
Furious, Hani accuses the authorities of playing with children’s futures by multiplying “improvised measures”.
Ahmed Tessa, an education specialist and author of a book on the status of French in Algerian schools, agrees.
“Honestly, how can one expect proper language teaching when you recruit university graduates and send them into classrooms without at least two years of initial training – the minimum needed to teach well,” he told MEE.
Tessa argues that the new measure is “an ideological rather than pedagogical decision”.
Rejection of French heritage
The government’s announcement has triggered a broader debate over the legacy of French colonialism and Algeria’s relationship with France.
El Hachemi Djaaboub is a former minister from the Movement of Society for Peace (MSP), an offshoot of the Muslim Brotherhood, and known for his critical stance towards French influence. He argued that postponing French teaching is an act of sovereignty and accused Tessa of serving foreign interests.
The downgrading of the language of Molière in Algerian schools comes amid persistent diplomatic tensions between Algeria and France, fuelled in 2024 by President Emmanuel Macron’s support for Morocco’s claim of sovereignty over Western Sahara. The former Spanish colony, largely controlled by Morocco, is also claimed by the Sahrawi independence‑seeking Polisario Front, which is backed by Algiers.
‘We must put an end to France’s colonial legacy. Even President Macron speaks more English than French abroad’
– Abderrazak Makri, former president of the MSP
In France, some right‑wing and far‑right media have criticised Algiers’s decision to promote English, seeing it as an attempt to reduce the French cultural and political influence in the country.
In Algeria, social‑media conversations suggest that part of the population now regards French as a colonial leftover to be cast aside. Some even call for the complete suppression of French teaching.
For Abderrazak Makri, former president of the MSP, the change must be radical.
“We must put an end to France’s colonial legacy. Even President Macron speaks more English than French abroad. Today, English is the universal language; it is time for Algerians to adopt it and align themselves with great nations,” he said in 2022.
A few political parties, such as the Rally for Culture and Democracy and the Workers’ Party, are, by contrast, opposed to downgrading French, arguing that language teaching should be complementary and kept away from ideological quarrels.
Analysing positions hostile to the presence of French in education, Sebaa speaks of “visceral, ideologically driven postures”.
“Some of their promoters studied in English-speaking universities and consider English an international language, whereas French symbolises colonial domination,” he told MEE.
Author of a book analysing Algerians’ relationship with French, the sociologist rejects the notion that colonial schooling significantly contributed to the “francisation” of Algerians, whether culturally or linguistically. At the time, he noted, only a small share of Algerian children were in school – less than 20 percent at the time of independence.
Sebaa added that Algerians who lived under colonial rule saw French rather as a means of social advancement and cultural promotion. He pointed to the emergence, at the time, of Algerian literature written in French as evidence. One of its most emblematic figures, Kateb Yacine, famously described French as “a war booty”.
According to Tessa, this heritage is now threatened. He warns that downgrading French will have harmful consequences in many respects. He particularly fears a drying‑up of cultural production in French and an erasure of collective memory expressed in this language.
The educator also warns of Algeria’s distancing from its francophone geopolitical environment and a rupture with millions of children and grandchildren of the diaspora, living mostly in France and Quebec.
Currently, according to the International Organisation of La Francophonie, Algeria is the third‑largest francophone country in the world, behind France and the Democratic Republic of Congo. About 15.6 million Algerians reportedly speak French – about 33 percent of the population."
https://www.middleeasteye.net/news/algeria-focus-english-schools-rekindles-debate-over-french-colonial-legacy
#metaglossia
#metaglossia_mundus
"Interpreter of the Deaf and Hard of Hearing - 7.0 hrs/day
Positively transforming learning and life outcomes for students through educational justice, equity, and excellence.
Qualifications
Valid Illinois Professional Educator License Approval as an Interpreter - Sign Language
Compensation
The hourly wage for this position is determined based on the Collective Bargaining Agreement with the Champaign Educational Support Professionals, which is available on our website. The District gives experienced employees up to 20 years of experience credit for service outside the District if it is directly related to their position. 100% of an employee's contributions to IMRF are currently paid by the Board of Education. The Board also pays 100% of employee-only health and term life insurance. There is a minimal employee cost for vision insurance. Dental is also available at competitive rates to interested employees. There are opportunities for employees to earn additional compensation during the regular school year and in the summer.
Application Deadline
This position will remain posted until the hiring process is complete; however, interviews may be conducted anytime after September 15, 2026, so interested applicants should submit their materials ASAP to ensure consideration.
More About Our Community
Champaign Unit 4 Schools serves over 10,000 PreK-Young Adult students in 18 facilities, which include twelve elementary schools, three middle schools, two high schools, one early childhood facility, and the Novak Academy. Located conveniently near Chicago, Indianapolis, and St. Louis, Champaign residents enjoy the entertainment and intellectual activity of a large city, without the higher costs of living and lengthy commute times often associated with busy city/suburban life. The University of Illinois provides the excitement of Big 10 sporting events, a thriving business environment, and world-class entertainment at State Farm Center and the Krannert Center for the Performing Arts. Champaign is characterized by a thriving downtown and campustown, shopping centers and boutiques, award-winning restaurants, neighborhood parks, live music venues, and summer festivals. We also share many museums, galleries, and theatres with the adjacent city of Urbana.
Want to Learn More?
If you want to learn more about this position or other career opportunities in Champaign Unit 4 Schools, please contact our Assistant Director of Human Resources & Employee Engagement, Ms. Nancy Whitehouse, at whitehna@u4sd.org. Please include your résumé."
https://www.k12jobspot.com/Job/5708871
#metaglossia #metaglossia_mundus
"Daily Kos Staff
Tech
AI hasn’t replaced these interpreters— but it has degraded their working conditions
by Capital & Main
FollowFollow "Capital & Main" to receive notifications about new pages on "Capital & Main".
Sunday, 13 September 2026 at 19:00 GMT+1
By Gabriel Thompson for Capital & Main
When Sara Ramirez started as an interpreter for Monterey, California-based LanguageLine Solutions, in October 2020, she saw it as “a regular company.” She worked part-time as a Spanish-language interpreter from her home in El Paso, Texas, translating by phone for clients that included the Federal Emergency Management Agency, banks and hospitals. If she encountered a problem, she emailed her supervisor. If she still had questions, they talked on the phone.
Things began to change quickly after LanguageLine introduced a new AI workforce management program in 2025, Ramirez said. Before, she had her break midway through her five-hour shift. Now it moved around, seemingly at random. Entire shifts were eliminated from her schedule with little notice. The workload also increased, giving her as few as 15 seconds between calls during busy hours, when previously she had a few minutes.
The AI program, called NiCE and made by an Israeli-American company with the same name, promises employers it will create “a more agile, efficient, and human workforce.” For many of the workers themselves, however, the experience has been negative. “NiCE is supposed to optimize the perfect workload,” said Karolina Yermak, a Russian interpreter at LanguageLine. “But you get the workload of two instead of one, without enough time to even take a sip of water between calls.”
Interpreters are often listed among the workers at the highest risk of replacement by AI. But interviews with a half dozen LanguageLine employees reveal a different phenomenon: AI is not so much eliminating as eroding the quality of their work. In a recent survey of 161 workers conducted by the Communication Workers of America, who are seeking to unionize the interpreters, a majority of LanguageLine workers reported high levels of burnout, insufficient time between calls and frequent back and neck pain.
“No matter how hard you try to stay sharp, you lose it, because there is only so much you can take,” said Yermak about the pace of work needed to meet AI metrics set by LanguageLine. “By the end of the day, your brain is so fried that you make mistakes. The intensity leads to massive burnout.”
A LanguageLine spokesperson who declined to be named said in an email that the rollout of AI software to forecast volume and create schedules had landed during a difficult stretch for the industry, and they have since adjusted and operations are “back to normal.” The spokesperson attributed workers’ complaints about erratic schedules primarily to federal budget cuts that have strained the resources of their clients, as well as an immigration crackdown that has left many immigrants afraid to access services.
“We acknowledge and regret the disruptions our interpreters have experienced and we deeply appreciate their patience and commitment,” the spokesperson wrote in a statement.
* * *
LanguageLine Solutions grew out of a project launched in 1982 by a San Jose police officer struggling to communicate with the city’s Vietnamese refugee population and a former U.S. Marine who had served as an interpreter in Vietnam. Today, that volunteer initiative has become the largest interpretive services company in the world, with particularly rapid growth during the last decade. Since LanguageLine was purchased in 2016 by France-based Teleperformance, the number of interpreters at the company has grown from 8,000 to 24,000, including nearly 3,000 U.S.-based employees whom Communication Workers of America is seeking to organize.
In 2025, the company reported $1.1 billion in revenue and a client list that includes hospitals, government agencies, schools, immigration courts and private companies. Many of the situations in which interpreters work are highly charged, demanding nuance, judgment and precision. One call could be for an immigrant seeking asylum, another for a patient receiving a sensitive medical diagnosis, a third for a frantic individual calling 911.
Even before its deployment of AI management software in 2025, LanguageLine faced scrutiny for its treatment of employees. In 2015, an investigation by the U.S. Department of Labor found that the company owed interpreters $1.47 million after failing to properly pay overtime and prevailing wage rates. And last year, the company settled a class action lawsuit for $3.725 million brought by interpreters who alleged that the company failed to pay minimum and overtime wages. (The settlement did not include an admission of wrongdoing by LanguageLine.)
Related | Police use of artificial intelligence grows as rules lag behind
Low wages remain a concern for interpreters. The LanguageLine spokesperson told Capital & Main that the average hourly wage of a full-time interpreter in the U.S. exceeded $20 an hour, but four LanguageLine workers with the highest possible level of certification told Capital & Main they earn less than $20 an hour.
It’s the working conditions, however, that have driven many interpreters to join the union drive. The introduction of NiCE, the AI-powered workforce software, has led to forced time off with little notice, workers said. “Sometimes they cut my hours, sometimes a full day,” said Aizo Nokes, a Spanish-language interpreter in Florida. “Reduced hours means a reduced paycheck and reduced security,” she added.
Interpreters have also found their breaks between shifts moving around, seemingly at random. “They say it’s AI powered, and you can see that,” said Yermak. “It cuts up your shifts into strange, pointless segments that aren’t convenient for any live human being.” Instead of evenly spaced breaks, she said, she could have a lunch break and then another break just 45 minutes later.
Related | As AI use in schools grows, lawmakers and districts scramble to set up guardrails
Many interpreters also reported that the introduction of the AI management system had increased the number of calls they typically receive in a shift. Before, they usually had several minutes between calls, but the minimum time between calls had now been shortened to 15 seconds.
“It’s like they flipped a switch for me,” said Agnieszka Lamirowska, a Polish interpreter who had been with the company for a decade before being fired in August due to what she said the company identified as “performance issues.” She said her call volume has more than doubled in the past year, up to 50 calls a day, and she received a deluge of alerts when she wasn’t meeting her goals. (LanguageLine said that from January to June of 2026, the median time between calls for U.S. interpreters was 2 minutes and 15 seconds and the average time was 5 minutes and 43 seconds.)
The implementation of NiCE, according to workers, had also led to more intense surveillance of their calls with clients based on metrics they did not fully understand. Interpreters said they could be penalized for situations they couldn’t control, such as having a call dropped due to faulty company equipment or having a client hang up before they could say their closing remarks. “You wake up in the morning and you don’t want to log in, because your stomach hurts,” said Nonése Kissane, a Haitian Creole interpreter in Florida. “The stress is too much.”
* * *
Interpreters at LanguageLine expressed some optimism that AI would not replace them entirely. One factor in their favor: Customers want to talk to people. “At the beginning of calls, I’ve had people ask me, ‘Are you a robot? Are you a recording?’” said Ramirez. Customers are relieved to learn she’s human.
Madhurima Ganguly, a Bengali video interpreter in Michigan who mostly takes hospital calls, described the relief her bewildered clients express when she greets the patient in their native language. “You have to understand that I’m the person they connect with,” she said. “I am their sister.” Ganguly absorbs their anguish, too. She recounted interpreting for parents of a child in the neonatal intensive care unit who were learning their baby might not survive. “It’s heavy, very heavy,” she said, close to tears. “The nuances, the emotions, the empathy that we humans have for one another … that will be there for eternity. No amount of AI can ever do that.”
A second factor may also protect the jobs of LanguageLine interpreters: the potential legal risk to the company and their clients of an overreliance on AI. LanguageLine interpreters work in situations — like immigration court, emergency rooms or 911 dispatch calls — where even small errors can have outsized consequences. Full-fledged AI hallucinations could have disastrous consequences, interpreters said.
David Horsey/Tribune Content Agency
The centrality of humans to interpretation is a point on which LanguageLine’s senior management agrees with workers. In January, Simon Yoxon-Grant, the CEO of LanguageLine, wrote in a company blog post that, even in the age of AI, humans “bring judgment, cultural nuance, and the ability to navigate what cannot be reduced to patterns.” AI did not “diminish the human role” in interpretation, he wrote, but rather “elevated it by handling the mechanical so people could focus on the meaningful.”
Last year, the company announced on its website that, as AI interpretation had improved, it was introducing a pilot program that routed less complicated calls to AI interpreters, thereby “freeing up human interpreters for more demanding work.”
According to Lauren McFerran, former chair of the National Labor Relations Board and currently the executive director of the AFL-CIO Tech Institute, which analyzes the impact of new technologies on workers, this illustrates a broader pattern: What management perceives as increased efficiency, workers experience as intensification. As less emotionally and cognitively taxing calls are handled by AI, the more difficult work is left to humans.
“This is often a selling point,” said McFerran. “We’re going to delegate the easier tasks to this technology. It may be very efficient and productive from the employer’s perspective, but the workers’ lived experience with only doing the very hardest parts of their job in an intensely monitored fashion is a very different side of that story.”
At LanguageLine, the new AI pilot service provides interpretation in more than 10 languages to handle “high volume, low risk” situations, though the company spokesperson said that the pilot was small and that it was not designed to replace interpreter jobs. Multiple interpreters reported that they had already seen a decrease in some simpler calls, like setting up appointments or telling patients a refill was ready.
* * *
In April, the LanguageLine interpreters’ union drive received the backing of numerous New York City officials, including Comptroller Mark Levine, at a press conference held at City Hall. Workers have also formed an employee health and safety committee, and in July a delegation traveled to the company’s Monterey headquarters, asking for a meeting with management. To date, the company has not agreed to meet collectively with the workers.
One motive for the unionization drive is the workers’ desire to shape their working conditions, including how and where AI is used. “No one asked us about NiCE,” said Kissane. “They told us what they were doing and we had to figure it out.”
For interpreters like Kissane, workers, not AI or managers, are the ones who know what conditions they need to do their work. “Let us take breaks when we need a break,” she said. “We know when our back hurts. We know when we need a walk. We know when a call is so deep that we can’t take another one.”"
https://www.dailykos.com/stories/2026/9/13/800096420/tech/ai-hasnt-replaced-these-interpreters-but-it-has-degraded-their-working-conditions/
#metaglossia
#metaglossia_mundus
“Machine translation is still broken for most of the world’s languages”: Cohere builds non-reasoning for a reason
Cohere's North Small Translate beats DeepL and Google Translate on WMT26 across 50 languages — but commercial use requires a Model Vault license.
Developers can download the weights for noncommercial use under CC BY-NC 4.0. Cohere offers commercially licensed deployment through Model Vault, which is a Cohere-managed inference environment. Cohere positions the model as part of its sovereign AI strategy, aimed at organizations that want greater control over where their models run and how their data is handled.
North Small Translate builds on Cohere’s multilingual and translation lineage, which includes its Tiny Aya and Command A Translate model families. The company claims North Small Translate outperforms “similarly sized open-weight models” under 1T parameters, as well as API-based translation models in various dimensions of machine translation on average.
Cohere co-founder Nick Frosst tells The New Stack that the model’s efficiency draws from the fact that it is non-reasoning, i.e., it relies on learned statistical patterns without a step-by-step logic process, which means it uses fewer tokens.
Machine translation is still broken for most of the world’s languages
“We spent nine years scaling an architecture invented to fix translation, and machine translation is still broken for most of the world’s languages,” Frosst says. “General-purpose models get you most of the way and then stop. The next phase of enterprise AI in this space is smaller, more specialized, and runs inside your own walls.”
“…machine translation is still broken for most of the world’s languages.”
In Cohere’s reported evaluation using WMT26 benchmarks, the company states that North Small Translate leads with a WMT26 All Languages benchmark score of 83.60, compared with 81.56 for Qwen 3.5 397B A17B, 76.50 for GLM 5.2 FP8, 81.37 for DeepL NextGen, 79.46 for Gemma 4 31B (on), and 68.20 for Google Translate.
With its mixture-of-experts architecture and 218 billion total parameters, with 25 billion active. Cohere points to North Small Translate’s smaller compute & memory footprint than other models. Some model-to-model comparisons in this space aren’t fully substantiable, since not every vendor discloses parameter counts.
With current solutions, long documents start to fall apart
“Machine translation allows documents to be translated from one language to another automatically. With current solutions, long documents start to fall apart,” Frosst says. “Google Translate scores 21.3 on our long-context test, Gemma 4 31B 19.4; we score 48.9. That’s [for example] a safety manual that reads fine on page one… and has drifted by page ten. The other risk is where the text goes. Once you push HR policies or regulated documents through a third-party API, that data has left your building, and necessarily that means your control over it is diminished.”
“The risk [in machine translation] is where the text goes. Once you push HR policies or regulated documents through a third-party API, that data has left your building and necessarily that means your control over it is diminished.”
Explaining why the model offers “stronger translation performance” across complex enterprise translation tasks, Frosst says the model can support work spanning “a high volume” of sensitive documents.
As well as its 50 languages (32 ‘high-resource’ languages + 18 others), the Cohere team explains that the model also supports translation-workflow-focused capabilities, such as structured translations (i.e., Markdown or JSON documents), instruction following (i.e., recommended tone & format), and terminology guides (i.e., providing specific vocabulary to use in the translation), all as part of the model.
“North Small Translate works with a multi-pass workflow,” explains Frosst. “The model translates, reviews its own output, finds errors, and fixes them – and this is the same loop we used in training. We ship both because standard is one pass and built for volume, while the agentic [version] spends more tokens for 84.36 against 83.60 on WMT26. That difference ends up being worth it when the document is a contract or a safety procedure, for instance, but in other cases you’d rather optimize for efficiency.”
“The model translates, reviews its own output, finds errors and fixes them.”
Model ‘steerability’ drives suggesting language tone and formatting
This model uses the same architecture as prior Cohere models but improves performance through post-training advances, including reinforcement learning and new datasets, specifically for machine translation tasks.
Frosst concludes that, across the translation model marketplace, generative machine translation models offer the highest quality and steerability (i.e., suggesting tone, formatting, etc.) but typically cost much more than Neural Machine Translation (NMT) models commonly used in commercial use cases.
North Small Translate was developed in partnership with RWS, an AI solutions company pioneering in language technology and services. Collaboration with RWS, specifically with its Language Weaver research and science teams along with its language experts, helped shape the model’s real-world translation performance throughout development.
As noted above, developers can access the weights free of charge for non-commercial use in three quantizations. There is also a Hugging Face Space and an API for those who lack the required hardware. "
Sep 13th, 2026
Adrian Bridgwater
https://thenewstack.io/cohere-north-translate-sovereignty/
#metaglossia
#metaglossia_mundus
"Booker-winning translator Deepa Bhasthi on why she has ‘nothing kind to say about AI’
Bhasthi discusses preserving the soul, musicality and rhythm of a language, and why a translation should not erase its cultural roots.
International Booker Prize-winning translator Deepa Bhasthi is in Kathmandu as one of the mentors at a three-day literary translation workshop at Kathmandu University’s Dhulikhel campus.
The workshop, organised jointly by Kathmandu University’s Department of Languages and Mass Communication and the University of Chicago’s South Asian Literature in Translation (SALT), brought together writers and translators working with Nepali, Maithili, Tibetan and Nepal Bhasa.
Bhasthi became the first Indian translator to win the International Booker Prize in 2025 for her translation of Banu Mushtaq’s ‘Heart Lamp’, a collection of 12 short stories written over more than 30 years that chronicle the everyday lives and struggles of women in southern India.
In this conversation with Post’s Anish Ghimire, she talks about why a translation should retain the ‘soul’ and rhythm of its original language, the importance of reading more South Asian literature, and why she has ‘nothing kind to say about AI’.
‘Heart Lamp’ became the first collection translated from Kannada to win the International Booker. What was particularly difficult about translating Mushtaq, not just linguistically, but culturally and emotionally?
Every translation project comes with its own challenges and constraints. The characters in Mushtaq’s stories are Muslim, but the problems they face are not restricted to the Muslim community. They are the experiences of women living under patriarchy around the world. That relatability is what made it easy for me to understand these characters.
Culturally, though, they were very different from my lived experiences, so I did a lot of research. I like to call it ‘method translation’ because I really tried to get into the culture to the best of my abilities. Obviously, it’s not possible to live like these characters, but I watched a lot of Urdu serials and films, and listened to Sufi-infused music to set the mood.
Emotionally, I’m not a very emotional reader. Reading is very different from reading as a potential translator of that work. I’m not very sentimental in that way. I see it as a story and as language, so you can’t let that affect you.
I don’t think it would make for a good translation if you approached something too emotionally. You have to step back from being a fan of the work and look at it from a much more critical perspective.
You speak about “translating with an accent” rather than making a text sound completely natural in English. What does an accent allow a translation to retain?
The soul of the original language—the soul, the musicality, the rhythm of that particular language.
Because, there is no such thing as standard English. I don’t think it even exists. Also, I don’t think that’s the point of translation at all. Otherwise, it just becomes a very clinical, very literal, very dry work without the essence.
You need a soul to have a life. Languages have these souls. Obviously, we can’t carry them over to another linguistic culture perfectly, but we can try to do so to the best of our ability.
Banu Mushtaq (left), author of ‘Heart Lamp’, and translator Deepa Bhasthi. Photo: Courtesy of The Booker Prize Foundation
Adding to that, what do you think should remain untranslated in a translation?
I don’t think there’s a one-rule-fits-all idea in translation. It depends on the project: what to retain, what to leave untranslated. I think it’s a word-by-word choice that one has to make: whether something should remain untranslated or be transliterated.
Who is the translator ultimately responsible to—the original writer, the original language, the reader, or the text itself? And what happens when those responsibilities conflict?
I think the conflict is always going to be there, especially in the case of Kannada and English. They are very different languages, with almost nothing in common, so the conflict will always exist. But again, I think some of the best literature exists within these conflicts.
I can’t really speak for all translators, but for me, I am always led by language. The language is important to me. I am not taken in by the fame of a writer.
For me, the question is: how best can I convey the linguistic culture of Kannada into English? That’s my area of interest.
Nepal has a rich literary tradition in Nepali and many other languages, but relatively little Nepali literature reaches international readers through translation. What do you think is required for more South Asian literatures to travel across linguistic borders, not just into English, but between each other?
There just have to be more translations.
We’ve always had a culture of translation in South Asia, if you want to look at it that way. Because we’re all thinking in two or three languages. We’re always translating between these languages, whether we’re at home, at work, or with friends. We speak different languages in our heads. We live in different languages.
As AI becomes increasingly capable of translating languages, what do you think it still struggles to understand about a literary text? And what might be lost when we allow a machine to make those choices for us?
I have nothing kind to say about AI.
I don’t think AI is anywhere close to being capable of translating literary fiction. When AI is capable of shedding tears over sorrow, laughing at a silly joke, and, you know, being human, then I’ll decide to worry.
If an AI produces a translation that readers cannot distinguish from a human translation, does it matter who translated the book?
That’s a very interesting question.
It might not matter to the person who is giving the prompt, but I think it is an insult to hundreds of thousands of years of human evolution. It is insulting to a person’s intelligence and critical-thinking abilities to take something a machine has produced and try to pass it off as something a human has done.
Am I completely against AI as a technology? Yes, for sure. Am I against it in creative fields? Oh my goodness, yes!
AI has its uses in some fields—science, medical science, and so on—so please use it to do whatever it can there.
But we don’t need this stupid ’80s social media trend, or to feed all these big tech companies at our expense. I mean, come on, at what cost? We are all seeing what is happening because of these big tech companies.
What do you think young translators worry about too much—and what should they be worrying about instead?
I don’t think young translators worry enough about critical thinking. Everyone is so geared towards doing everything quicker. I think translators don’t read enough. I always say you have to read more than you write, and listen more than you talk.
And we should read more South Asian literature. We’re very happy to read translations from French, Italian and German, which we should, but we also need to read stories from our own part of the world. That’s the only way we’ll have the confidence to tell the world, ‘Hey, our story is my favourite.’
Talking about books, can you give me your 3 quick book recommendations?
I’m currently reading ‘Brotherless Nights’ by V V Ganeshananthan, which is incredible.
Anything by Claire Keegan, especially ‘Foster’.
And thirdly, I’d recommend a Kannada book, ‘Karimayi’, written by Chandrashekhara Kambara and translated by Krishna Manavalli."
Anish Ghimire
Published at : September 14, 2026Updated at : September 14, 2026 07:13
https://kathmandupost.com/art-culture/2026/09/14/booker-winning-translator-deepa-bhasthi-on-why-she-has-nothing-kind-to-say-about-ai
#metaglossia
#metaglossia_mundus
Stimuler la traduction et renforcer le rôle de la littérature dans la diffusion de la culture et du savoir
La Commission de la Littérature, de l’Édition et de la Traduction (LPTC) a annoncé l’ouverture des inscriptions aux maisons d’édition saoudiennes et internationales souhaitant participer à la Foire internationale du livre de Riyad 2026, prévue du 10 au 19 décembre.
Riyad, 14 septembre 2026, SPA -- La Commission de la Littérature, de l’Édition et de la Traduction (LPTC) a annoncé l’ouverture des inscriptions aux maisons d’édition saoudiennes et internationales souhaitant participer à la Foire internationale du livre de Riyad 2026, prévue du 10 au 19 décembre.
La LPTC a invité les maisons d’édition et les institutions intéressées à soumettre leurs demandes d’inscription via le lien suivant : https://engage.moc.gov.sa/reg_forms/tracks/326/?lang=en.
La Foire internationale du livre de Riyad constitue un grand rendez-vous culturel annuel. Elle réunit des professionnels de la littérature, de l’édition et de la traduction venus du monde entier, en vue de soutenir l’industrie du livre, de stimuler la traduction et de renforcer le rôle de la littérature dans la diffusion de la culture et du savoir. Elle offre également aux éditeurs une plateforme pour présenter leurs nouveautés au public.
Cette édition comprendra un programme culturel riche en tables rondes, ateliers et conférences, avec la participation d’écrivains, d’universitaires et d’experts saoudiens et internationaux. La Foire maintiendra également l’espace « Coin de l’auteur saoudien », destiné à permettre aux écrivains saoudiens de présenter et de promouvoir leurs ouvrages publiés à compte d’auteur.
Une attention particulière sera par ailleurs accordée à la littérature pour enfants et adolescents à travers le « Pavillon de l’enfant », qui proposera des programmes interactifs et des ateliers créatifs visant à stimuler l’imagination et à encourager la lecture dans un environnement éducatif ludique et sécurisé.
La Foire internationale du livre de Riyad constitue l’un des grands rendez-vous culturels du monde arabe, grâce à sa forte affluence, à la diversité de son programme et à la large participation des maisons d’édition saoudiennes et internationales, contribuant ainsi à renforcer la position du Royaume comme pôle culturel de premier plan sur la scène internationale.
-- SPA
09:09 Heure locale 06:09 GMT
0019
lundi 3 Rabi' al-Thani 1448 - 14 septembre 2026
https://www.spa.gov.sa/fr/N2675611
#metaglossia
#metaglossia_mundus
Obtenir la traduction de quelque 6000 anciens jugements de la Cour suprême
L'organisme Droits collectifs Québec poursuit son combat juridique pour obtenir la traduction de quelque 6 000 anciens jugements de la Cour suprême…
L’organisme Droits collectifs Québec (DCQ) et son directeur général, Étienne-Alexis Boucher, se tournent vers la Cour d’appel fédérale pour obliger la Cour suprême du Canada à traduire en français ses documents d'avant 1970. L'ancien député du Parti québécois et son organisme ont porté en appel, le 26 août, le jugement de première instance qui leur était défavorable. Ils estiment que la décision de la juge Denise A. LeBlanc est affectée par un certain nombre d’erreurs de droit et d’interprétation juridique déterminantes justifiant de faire appel, selon les documents transmis à Radio-Canada.
« On ne pense pas que c’est un bon jugement pour un pays bilingue », souligne Étienne-Alexis Boucher, directeur général de Droits collectifs Québec.
« On pense qu’on a des motifs raisonnables et pertinents de porter appel d’une décision qui, à nos yeux, non seulement ne respecte pas la jurisprudence établie - notamment par l’arrêt Beaulac, qui vise à interpréter de manière très large et généreuse les droits linguistiques, notamment ceux des francophones -, mais aussi parce qu’on ne voit pas comment ce jugement permet au Canada de tendre vers une égalité réelle entre tous ses citoyens, peu importe la langue officielle qu’ils parlent », explique M. Boucher en entrevue à Radio-Canada.
6000 décisions à traduire
DCQ et son directeur général souhaitent voir traduites en français quelque 6000 décisions rendues avant l'entrée en vigueur, en 1969, de la Loi sur les langues officielles (LLO). Avant cette loi, le plus haut tribunal du pays n'était pas tenu de publier ses jugements dans les deux langues officielles du Canada. La plupart de ses décisions étaient rendues en anglais, même lorsqu’il s’agissait de causes ayant rapport avec le Québec.
En juin dernier, la juge LeBlanc s'était rangée derrière les arguments du Bureau du registraire de la Cour Suprême du Canada. Elle avait conclu que les décisions historiques de la CSC ne constituent ni des services ni des communications au public soumises à l’obligation de traduction prévue par (...) la LLO. Selon la magistrate, l'obligation de respecter les deux langues officielles du Canada ne s'applique qu’au site web de la Cour suprême, et non pas au contenu des décisions à proprement parler.
Le Bureau du registraire est l'instance visée par la plainte, puisque la Cour suprême du Canada elle-même est protégée contre ce genre de recours en vertu du principe de l’indépendance des tribunaux. Invité à réagir, le Bureau a refusé de commenter le dossier comme il s’agit d’une affaire actuellement devant les tribunaux.
Des traductions, à quel coût?
La démarche judiciaire entreprise a déjà permis d'obtenir une victoire, souligne M. Boucher. En 2025, la Cour suprême du Canada s'est engagée, pour les 150 ans du tribunal, à traduire en français 24 des principaux jugements rendus avant 1970. Mais en ne traduisant pas l'ensemble des 6000 décisions, Droits collectifs Québec et M. Boucher estiment que la Cour suprême du Canada continue d'enfreindre la Loi sur les langues officielles, ce qu'avait aussi conclu le Commissaire aux langues officielles (CLO) dans deux rapports, en octobre 2021 et en septembre 2024. Par le passé, le juge en chef de la Cour suprême du Canada, Richard Wagner, avait toutefois affirmé qu'une telle opération serait beaucoup trop onéreuse.
De plus, selon lui, l'utilité de traduire ces décisions serait minime, la plupart d'entre elles n'ayant plus vraiment de portée dans le système juridique moderne. Deux arguments que conteste M. Boucher. Le manque de ressources (...) ce n’est pas une excuse pour brimer les droits fondamentaux. Plusieurs cours le disent à travers le Canada, dont la Cour suprême, souligne l'ancien député péquiste. Il rappelle également que de très nombreux jugements rendus avant 1969 font partie de la jurisprudence et sont encore cités par les parties qui veulent remporter des litiges devant les tribunaux.
« Ne serait-ce que pour permettre à la Cour (suprême du Canada) de donner l’exemple, (...) nous, on pense d’un nécessaire absolu que la Cour suprême procède à la traduction desdits jugements », mentionne Étienne-Alexis Boucher.
Pour lui, c'est une question d'égalité pour l'ensemble des juristes d'expression francophones de pouvoir disposer des mêmes outils, des mêmes avantages que leurs collègues anglophones. L'avocat Gabriel Poliquin devant la Faculté de droit de l'Université d'Ottawa.
Une déclaration que confirme l'avocat constitutionnaliste Gabriel Poliquin, qui est récemment intervenu dans l'affaire de la nomination de la lieutenante-gouverneure du Nouveau-Brunswick.
« Les seuls jugements qui étaient vraiment pertinents, c'étaient des jugements d'avant 1970, des jugements du Conseil privé en Angleterre (...) disponibles seulement en anglais », raconte celui qui est également le directeur du Programme de certification de common law en français de l'Université d'Ottawa.
Les appelants demandent que toutes ces décisions soient traduites dans un délai de 10 ans... ou plus si nécessaire.
« DCQ n’est pas dogmatique à un point tel qu’on imposerait à la Cour suprême un défi qui est insurmontable. (...) Je pense qu’on pourrait simplement pousser plus loin les travaux déjà entamés et faire une hiérarchie de ces travaux-là. Est-ce que ça pourrait prendre 10 ans, 15 ans, 20 ans? Ce que nous demandons, c’est un plan de match ».
Pour Me Poliquin, l'appel qui a été déposé est d'autant plus pertinent que plusieurs points restent à clarifier après le premier jugement.
« La question juridique, c'est une question vraiment de noir et blanc : soit il existe une obligation (de traduire), soit elle n'existe pas », mentionne Me Gabriel Poliquin, avocat constitutionnaliste Mais pour lui, même si la Cour d'appel fédérale donne tort aux appelants, un compromis pourrait encore être trouvé.
« On pourrait arriver à un compromis (...) sur le plan pratique d'au moins traduire les décisions qui sont d'intérêt public ».
L'audience devant la Cour d'appel fédérale devrait se tenir à Montréal à une date qui n'a pas encore été révélée."
Radio Canada
2026-09-02 10:30:29
https://droit-inc.com/conseils-carriere/nouvelles/traduction-a-la-cour-supreme-la-saga-judiciaire-nest-pas-terminee
#metaglossia
#metaglossia_mundus
|
Scoop.it!
Chinese technology giant Baidu has overtaken Google and Microsoft in an artificial intelligence competition designed to test how well a machine can understand human language.
Baidu, which is often referred to as China‘s Google, achieved the highest ever score in the General Language Understanding Evaluation (Glue) – widely considered to be the benchmark for AI language understanding.
The firm’s Ernie (Enhanced Representation through kNowledge IntEgration) model became the first to score above 90 on the test, topping a leaderboard dominated by US tech firms and universities.
The feat also makes it one of only 10 AI systems to surpass the average human score of 87.1 on the GLUE benchmark.
Ernie used a similar method to Google’s Bert (Bidirectional Encoder Representations from Transformers) model, which transformed natural-language understanding for AI when it was created last year.
Gadget and tech news: In pictures
Both Bert and Ernie – named after Sesame Street characters – interpret meaning by examining the words that appear both before and after a word in a sentence in order to fully establish context.
By first developing the Ernie model using the Chinese language and then using it for English words, Baidu researchers realised it made the algorithm even stronger at understanding English.
Read more
"When we first started this work, we were thinking specifically about certain characteristics of the Chinese language. But we quickly discovered that it was applicable beyond that," Hao Tian, chief architect of Baidu Research, told MIT Technology Review, who first reported on the research.
Baidu is already using the model to improve results for its search engine and make its AI assistant Xiao Du more accurate.
A paper detailing how Ernie was trained for the language test will be presented at the Association for the Advancement of Artificial Intelligence conference next year.