Metaglossia: The Translation World

"Errors in automated translation systems used by California courts are raising concerns that residents with limited English proficiency may be denied meaningful access to justice.

The issue is particularly relevant to Korean Americans, as Korean remains one of the most frequently requested languages for court interpretation in the state, underscoring the need for accurate and reliable interpretation and translation services.

Thirty-five legal and community organizations, including California Rural Legal Assistance, Inc. (CRLA), the Legal Aid Foundation of Los Angeles (LAFLA) and the Mixteco Indígena Community Organizing Project (MICOP), recently called on the Judicial Council of California to suspend the use of automated translation in state courts and strengthen language access services through qualified interpreters and translators.

In a joint public comment submitted earlier this month to the Language Access Subcommittee of the Judicial Council’s Advisory Committee on Providing Access and Fairness, the organizations said courts have failed to fully meet language access standards, leaving many residents with limited English proficiency unable to fully understand or participate in critical legal proceedings.

At a public hearing before the subcommittee, residents also voiced repeated calls for interpretation services provided by trained human interpreters.

Korean ranks eighth in court interpretation demand
The scale of language assistance needed in California courts is substantial, according to the Judicial Council’s 2025 Language Need and Interpreter Use Study, released last year.

From fiscal year 2020-21 through fiscal year 2023-24, California’s 58 superior courts recorded a total of 2,534,207 interpretations.

Korean accounted for 14,410 interpretations during the four-year period, ranking eighth among all languages, behind Spanish, Mandarin, Vietnamese, American Sign Language, Punjabi, Cantonese and Arabic. Korean is also among the languages for which the Judicial Council maintains a certification program for professional court interpreters.

Demand for Korean-language interpretation was heavily concentrated in the Los Angeles area. According to the report, 70.8% of all Korean interpretations took place in Region 1, which includes Los Angeles, San Luis Obispo and Santa Barbara counties.

Region 1 recorded 10,214 Korean-language interpretations during the study period, representing 1.7% of all interpretations in the region. Excluding Spanish, Korean had the second-highest number of interpretations after Mandarin.

Korean interpretation particularly significant in civil cases
The report also found relatively high demand for Korean interpretation in civil cases.

Statewide, Korean was used in 5,315 civil-case interpretations, accounting for 3.7% of all civil interpretations during the study period. Excluding Spanish, Korean ranked second only to Mandarin, which accounted for 6.0%.

Civil cases can include matters such as eviction, domestic violence and civil harassment, proceedings that can directly affect a person’s housing, safety, property and livelihood.

Language assistance needs are significant across California as a whole. As of 2022, about 16.29 million residents age 5 and older, or 43.9%, spoke a language other than English at home. About 6.36 million people, or 17.1%, were classified as limited-English-proficient, meaning they spoke English less than “very well.”

At the household level, 8.4% of California households were classified as limited-English-speaking households. In Los Angeles County, home to a large Korean American population, the share was 12.0%, well above the statewide average.

The report also noted that Korean speakers, along with Chinese and Vietnamese speakers, continue to have relatively high levels of limited English proficiency, indicating an ongoing need for targeted language assistance.

‘DUI’ translated as ‘IUD’
Despite the high demand for language services, legal advocates say courts are increasingly relying on machine translation without sufficient review by qualified translators or interpreters.

In 2015, the Judicial Council adopted the Strategic Plan for Language Access in the California Courts, establishing 75 recommendations intended to ensure equal access to court services for people with limited English proficiency.

According to the Judicial Council’s 2025 report, more than 60 of those recommendations have been completed, while several remain in progress. The report also states that the council is developing training and guidance for court staff on the appropriate use of machine translation tools.

The organizations that submitted the joint public comment, however, said key language access measures remain inadequately implemented and criticized courts for using machine translation since 2020 without sufficient oversight by qualified human translators and interpreters.

Their comments included examples of automated translation errors drawn from actual court data.

In one case, instructions from a court staff member were translated as, “Sting the judge to reduce a fever,” a phrase bearing no resemblance to the intended meaning.

In another case, “DUI,” referring to driving under the influence, was translated as “IUD,” the abbreviation for an intrauterine contraceptive device.

The organizations warned that such errors can have consequences far beyond inconvenience. Court users may miss hearings or filing deadlines, lose opportunities to reduce fines or avoid points on their driver’s licenses, or make important legal decisions without fully understanding their cases.

Joann H. Lee, the attorney overseeing language access at LAFLA, said California, as the nation’s most linguistically diverse state and home to the country’s largest court system, has a responsibility to ensure that language does not become a barrier to justice.

“For many people, court proceedings involve some of the most consequential moments of their lives,” Lee said. “Yet too often, they are left unable to fully understand or participate in decisions that affect their families, housing, safety and livelihoods.”

Aging and limited interpreter workforce
Securing enough qualified court interpreters is another growing concern.

As of March 2025, 1,856 certified and registered court interpreters were listed with the Judicial Council, covering 114 spoken languages and American Sign Language. However, 32% of interpreters on the master list were over age 65, highlighting concerns about an aging workforce.

As of 2024, there were 25 certified Korean court interpreters in California. Korean, meanwhile, remains one of the 10 most frequently interpreted spoken languages in the state court system.

Norma Ventura, Indigenous Justice Program Director at CRLA, said safeguards are essential as courts and other public agencies expand their use of machine translation and other AI technologies.

“As we see an increase in the use of machine translation tools and other AI technologies in courts and other public agencies, it is imperative that these entities implement safeguards that ensure accuracy,” Ventura said in the statement. “Safeguards such as qualified human review ensure equal access to services for all communities.”

Vanessa Terán, policy director at MICOP, said the Judicial Council’s 2015 language access plan represented important progress but has yet to fully deliver on its promise.

“More must be done to ensure that everyone, regardless of the language they speak, can fully understand and participate in proceedings that profoundly affect their lives,” Terán said.

The legal and community organizations are urging the Judicial Council to fully implement its existing language access plan before expanding automated translation and to stop using machine-generated translations that have not been reviewed by qualified professional interpreters or translators."
August 16, 2026
By Nicole Chang, originally published by American Community Media
Automated Translation Errors in California Courts Raise Concerns Over Korean Americans’ Access to Justice
https://asamnews.com/2026/08/16/translation-errors-california-courts-korean-americans-justice/
#metaglossia
#metaglossia_mundus

"Professions showing the biggest increase in job vacancies in recent weeks include authors, writers, translators and actors, according to new research.

In contrast, there was a fall in job adverts for secondary school teachers and undertakers, the Recruitment and Employment Confederation (REC) said.

Areas of the country with the biggest increase in job adverts generally included Milton Keynes, Westminster, Solihull and the Orkney Islands, said the report.

There were a total of 1,669,203 job adverts across the UK last month, a 7.7% increase from July last year, thanks to a big rise in hospitality and tourism jobs, recruiters said."
https://www.heraldscotland.com/news/26468674.uks-top-professions-biggest-rise-job-vacancies/
#metaglossia
#metaglossia_mundus

"Over the past decade progress has been made in expanding financial access across the globe. A key driver contributing to this growth has been the adoption of digital technologies that have helped overcome longstanding geographic and infrastructure barriers, bringing financial services to communities that were once difficult to reach.

Additionally, financial institutions are increasingly using alternative data to assess creditworthiness, enabling individuals with little or no traditional credit history to qualify for financial products that were previously out of reach.

Alongside these efforts to expand access, ensuring consumers have the knowledge and confidence to make informed financial decisions has become a key priority.

Financial literacy has become a growing focus for educators, employers, policymakers, and financial institutions alike. In the U.S. alone, more states now require financial literacy education before high school graduation, while employers are increasingly recognizing financial wellness as an essential employee benefit.

These initiatives are designed to help people understand concepts that are often intimidating even for native speakers. Interest rates, credit scores, debt management, investment planning, and retirement savings all come with terminology that can be difficult to navigate.

Now, imagine trying to understand those concepts in a language that isn't your own.

For millions of consumers, language is one of the greatest barriers to financial inclusion. In an industry built on trust, transparency, and informed decision-making, the inability to fully understand financial information can prevent consumers from engaging with products and services that could improve their financial well-being.

For financial institutions, addressing this barrier is both an inclusion initiative and a business opportunity. Financial institutions that fail to communicate effectively with multilingual communities risk overlooking one of the country's fastest-growing customer segments.

According to the Claritas 2025 Hispanic Market Report, Hispanic consumers now account for more than one-fifth of the U.S. population and are projected to drive approximately 80% of the country's population growth through 2031.

Some organizations are already responding. Old National Bank, for example, expanded its longstanding financial education program with Dinero Inteligente, providing educational resources for Spanish-speaking individuals and families.

Making information available in another language is only the beginning. Successful engagement depends on more than accurate translation. Customers are far more likely to trust and act on information that feels natural, culturally relevant, and designed with their experiences in mind. A message that is linguistically correct but culturally disconnected often fails to resonate. This is where localization becomes essential.

While translation focuses on converting words from one language into another, localization adapts content to reflect the cultural expectations, communication styles, and user experiences of a specific audience. It ensures that financial information is not only understood but also meaningful and approachable.

For financial institutions, localization represents an opportunity to advance both inclusion and growth. Research shows that organizations investing in localization are 1.8 times more likely to increase revenue and 1.5 times more likely to improve profitability. By creating experiences that are genuinely accessible, institutions can strengthen customer relationships while reaching underserved markets that have historically been overlooked.

Whether it's explaining lending terms, account fees, investment options, or retirement planning, clear and culturally relevant communication helps consumers make informed decisions with confidence. That confidence is the foundation of trust, and trust is the foundation of lasting financial relationships.

The financial services industry has made important progress toward greater inclusion, but future gains will depend on removing barriers that technology alone cannot solve. Expanding access is only part of the equation. Ensuring people can fully understand and engage with financial services — in their preferred language and within their cultural context — is what ultimately elevates access into opportunity."
https://financialit.net/blog/financialinclusion/beyond-translation-missing-link-financial-inclusion
#metaglossia
#metaglossia_mundus

"Walking with a friend in Bruges' Grote Markt. Credit: Anika Austvold
When I moved from the American Midwest to Brussels for the summer, I expected to encounter cultural differences. What I didn't expect was how often conversations would begin with assumptions.

Sometimes, those assumptions appeared the moment people heard my accent. I've been stopped and asked how I feel about US President Donald Trump before even introducing myself.

More than once, I've found myself explaining that no, not every American owns a gun, and yes, many of us travel extensively. Those conversations have made me realise that Europeans often hold simplified ideas about life in the United States.

At the same time, living in Brussels has highlighted some of the assumptions Americans commonly make about Europe.

Despite sharing centuries of history and consuming each other's media every day, Americans and Europeans often understand one another through misconceptions rather than reality. After several weeks living across the Atlantic, I've found that some of the most interesting conversations happen when those assumptions are challenged.

One continent, one country
After telling people in Brussels that I'm from Minnesota, I've often found myself explaining that life there looks very different from life in New York, Los Angeles or Washington DC.

Many Europeans I have spoken with understand that the United States is vast, but American culture is often represented abroad through a handful of major cities.

In reality, food, accents, political attitudes and social norms vary dramatically across the country. A Minnesotan, a Texan and a Floridian may all identify as American, but their experiences can be vastly different.

Living in Brussels has also made me realise how often Americans make similar assumptions about Europe.

Back home, Europe is frequently discussed as if it were a single cultural entity rather than a continent of distinct countries, cultures and political systems. Yet after spending time here, it becomes clear how much can differ from one country to the next, and sometimes even within the same country.

In Belgium alone, language, politics and culture vary significantly across the country. The closer you look, the harder it becomes to treat either Europe or the United States as one uniform culture.

Americans don't travel
Of all the assumptions I've encountered since arriving in Brussels, this is probably the one I've heard most often.

Many Europeans are surprised to learn how much Americans travel within their own country. While it's true that Americans are less likely than Europeans to hold passports and often take fewer international trips, geography helps explain why.

Travelling to another country from the US is rarely as simple as hopping on a train for the weekend. A flight from New York to Los Angeles takes roughly as long as a flight from Brussels to destinations in the Middle East, such as Egypt or Jordan.

The reality is not that Americans travel less. More often, we simply travel differently.

Departure hall in Brussels Airport. Credit: Belga/Nicolas Maeterlinck

Cost is another factor. While budget airlines make international travel relatively affordable within Europe, crossing the Atlantic often requires more planning, time and money.

Living in Brussels has also challenged one of the assumptions that I, like many Americans, grew up with: that Europe is small.

While crossing borders may be easier, travelling from Stockholm to Lisbon is farther than travelling from New York to Miami. Europe may be more connected than the United States, but isn't compact.

Divided America, united Europe
One of the first things people notice about me in Brussels is my accent. More than once, that has immediately led to questions about Trump, American elections and political polarisation.

The interest is understandable. For many Europeans, their understanding of the United States comes through political headlines. Presidential elections and partisan battles often dominate international coverage, creating the impression that Americans fall neatly into two groups: red (Republicans) and blue (Democrats).

While America's political system has become increasingly polarised, many Americans themselves are harder to categorise. Someone might support abortion rights while favouring stricter immigration policies. Another might support government-funded healthcare while opposing certain taxes. The reality is often far more nuanced than the headlines suggest.

Credit: Belga / David Stockman

Living in Brussels has also made me more aware of a misconception common among Americans. Back home, the European Union can be misconstrued as a "United States of Europe": a collection of countries moving in the same political direction under a common government.

But after spending time here, I've realised that Europe is far from politically uniform. Beneath the 12 stars on the EU flag are 27 governments with their own priorities, interests and political pressures. From migration and defence spending to energy policy and economic reform, member states disagree on many major issues.

Living to work
Before moving to Brussels, one of the stereotypes I heard most often about Europe was that people barely work.

The image is familiar to many Americans: long lunches, month-long summer holidays and generous benefits. It's easy to come away with the impression that Europeans have somehow cracked the code to a stress-free life.

But after a few weeks working in Brussels, that perception feels overly simplistic. On any weekday morning, it's hard to miss the thousands of people heading to offices, meetings and institutions across the city. Work is hardly absent from daily life.

Credit: Canva

Europeans, meanwhile, often view Americans as workaholics who define themselves entirely through their careers. While Americans generally receive fewer days off and often work longer hours, that doesn't necessarily mean they value work more than family, hobbies or leisure. In many cases, those choices reflect economic realities rather than personal preference.

The difference I've noticed is often less about whether people work hard and more about how work fits into their lives. In Europe, there seems to be a stronger emphasis on protecting time outside of work, while in the US, careers are often more closely tied to personal identity.

Armed America
After several weeks in Brussels, I've noticed a pattern: conversations about America often circle back to guns.

The perception is understandable. For many people outside the US, their exposure to America comes through headlines about mass shootings, debates over gun rights and stories about firearms. Those stories can create the impression that guns are a defining feature of everyday American life.

While firearms are far more common in the US than in most European countries, the reality is more nuanced. Many Americans have never owned a gun, and experiences with firearms vary dramatically depending on where someone lives, how they were raised and the communities in which they live.

English-speaking Europe
Living in Brussels has also made me more aware of a misconception many Americans have about Europe: that everyone speaks English.

Before arriving, I knew Europe was multilingual in theory. In practice, however, it has been striking to hear conversations shift seamlessly between French, Dutch, English and countless other languages throughout this city.

In many parts of Brussels, it is entirely possible to get by in English. But that accessibility can create the false impression that English is everyone's primary language.

Credit: Belga

Like many of the assumptions Americans and Europeans make about each other, some stereotypes contain a grain of truth.

During my second week in Brussels, I was buying groceries when someone asked me whether Americans actually cook.

The question made me laugh. There's a reason people associate America with fast food and convenience culture. But standing in a Delhaize with ingredients for dinner in my basket, I couldn't help but think about how easily a country's image can overshadow the everyday lives of the people who live there.

Many of the assumptions Americans and Europeans make about one another work the same way. We often judge entire societies by the most visible things, whether that's fast food, politics, holidays or headlines.

Living abroad has reminded me that the most interesting parts of a culture are often the ones that don't make it into the stereotypes, and those are often the parts you only discover once you leave your misconceptions behind."
Tuesday, 11 August 2026
By  Anika Austvold
https://www.brusselstimes.com/belgium/2185219/lost-in-translation-what-ive-learnt-as-an-american-in-brussels
#metaglossia
#metaglossia_mundus

languages form the core of personal identity, cultural preservation, and intergenerational connection

"Surrey festival celebrates language, culture and community

Surrey’s Holland Park became a celebration of language, culture and heritage on Aug. 9 as the 12th annual Mother Language Festival brought community members together from across the Lower Mainland.

Hosted by the Mother Language Lovers of the World Society (MLLWS), the festival ran from 3:30 p.m. to 7:30 p.m. It featured cultural performances, music, dance, art, literature, traditional food and community activities.

The annual celebration is inspired by International Mother Language Day, which normally takes place in late February, and aims to promote linguistic diversity, cultural understanding and preservation of Mother languages.

“The festival was a wonderful success, bringing together community members and performers from diverse cultural backgrounds to showcase their unique traditions and heritage,” said Islam.

For Islam, events like the Mother Language Festival are about more than just celebrating the languages spoken in Surrey and the Lower Mainland.

“More languages form the core of personal identity, cultural preservation, and intergenerational connection,” he explained.

Islam also said the festival aims to draw attention to languages that can be defined as being at risk of disappearing, such as Indigenous mother tongues and other minority dialects.

Islam explained that he believes protecting and revitalizing endangered languages is important to safeguarding human history, traditional knowledge and linguistic diversity.

He says the festival provides communities with a platform to share their languages and traditions while encouraging dialogue and connection between people of different backgrounds.

“By providing a platform for all voices, we promote mutual respect, inclusion, and cultural unity within our community,” said Islam.

Organizers say the festival attracts participants from across the Lower Mainland and beyond, offering an opportunity for people of all ages and cultures to celebrate their heritage while learning about the traditions of others.

Learn more about the MLLWS at motherlanguagelovers.com."
Published 2:30 pm Wednesday, August 12, 2026
By Grace McLeod
https://northdeltareporter.com/2026/08/12/surrey-festival-celebrates-language-culture-and-community/
#metaglossia
#metaglossia_mundus

"Google DeepMind a lancé un modèle de traduction de la langue des signes en texte, SL2T, dans deux produits Android grand public, la première fois que la technologie de reconnaissance de la langue des signes atteint une fonctionnalité de téléphone expédiée. Le modèle alimente la dictée de la langue des signes à texte dans Gboard et Live Transcribe sur les téléphones Pixel 11, avec une mise en ligne le 12 août 2026, pour la traduction de la langue des signes américaine en anglais.

La fonctionnalité fonctionne partout où un utilisateur taperait normalement. Un signeur sourd peut dicter une recherche Web, rédiger un message ou un document, ou interroger Gemini en signant à la caméra du téléphone au lieu de taper. Dans Live Transcribe, les signeurs peuvent répondre à une conversation signée plutôt que de taper des réponses. Google indique que les testeurs ont constaté que signer en langue des signes américaine était plus rapide et plus naturel que de taper en anglais.

SL2T est le premier modèle que Google décrit comme une avancée en termes de qualité et de généralité pour la traduction de la langue des signes. Il a été formé sur plus de 100 000 heures de données de signes, couvrant plus de 50 langues des signes, dont environ un quart des données en langue des signes américaine. La formation conjointe sur plusieurs langues, dialectes et niveaux de compétence a produit un modèle qui surpasse les systèmes à une langue dans les expériences de l’entreprise, selon l’annonce.

Ce que le modèle voit et comment il traduit
Le système ne traite pas les vidéos brutes du signeur. Un modèle sur appareil, MediaPipe Holistic, suit 130 points clés sur le visage, le corps et les mains, image par image, et seuls ces coordonnées géométriques quittent l’appareil pour la traduction. Le flux de caméra d’origine est immédiatement supprimé, une conception que Google présente comme une garantie de confidentialité.

À partir de ce flux de coordonnées, SL2T génère directement du texte en anglais, en sautant la couche d’annotation intermédiaire appelée gloses sur laquelle la plupart des systèmes de traduction de la langue des signes précédents reposaient. Les gloses attribuent des étiquettes fixes à des signes individuels, ce qui limite le vocabulaire et perd les marqueurs non manuels et les constructions spatiales qui portent un sens grammatical dans les langues des signes. La suppression de ce goulet d’étranglement permet à la qualité de la traduction d’évoluer avec les données de formation plutôt qu’avec un ensemble d’étiquettes construites à la main.

Les langues des signes sont des langues naturelles indépendantes avec leurs propres grammaires, et non des versions signées de l’anglais parlé. Cela rend la tâche une traduction automatique entre deux langues, plus un problème de vision par ordinateur difficile : le modèle doit suivre le mouvement simultané des mains, des bras, du torse, de la tête et du visage à des taux de trames élevés. Les tentatives antérieures de technologie de langue des signes, telles que les gants de capteur, ne pouvaient répondre à aucune de ces exigences.

Résultats des benchmarks et modèles d’erreur restants
Sur le benchmark FLEURS-ASL, qui mesure la traduction de la langue des signes américaine en anglais de langage complexe et abstrait enregistré dans des conditions de studio par des interprètes sourds certifiés, SL2T obtient un score de 70 BLEURT zero-shot, nettement supérieur aux résultats précédemment signalés, selon Google. L’entreprise signale également un score de 74 BLEURT sur une variante de signes à une main du benchmark et de 85 BLEURT sur PRESTO-ASL, un ensemble de données de phrases d’assistant signées à une tablette amarrée. Sur FSboard, un ensemble de données de signes manuels collecté à partir de signeurs sourds sur des appareils mobiles, le modèle atteint une précision d’appariement exact de 64 pour cent. Ce sont des chiffres rapportés par le fournisseur ; aucune évaluation indépendante n’a été publiée.

Google a publié des exemples côte à côte à partir de FLEURS-ASL montrant une sortie fluente aux côtés de modes d’échec identifiables. Le modèle remplace occasionnellement des signes rares et des signes manuels rapides, supprime les constructions passives et les dépeints de classeur, et perd la tension lorsque le contexte est manquant. Un exemple traduit « Ce volatile entièrement emplumé, à sang chaud » par « Cette créature est à sang chaud, mange du gris », une erreur de signe manuel substituant « gris » à « proie ».

Le modèle présente également un comportement résiduel d’hallucination, générant du texte lorsque personne ne signe, comme lorsque une deuxième personne entre dans le cadre ou que le signeur s’arrête. Google indique que la version de sortie a considérablement réduit ces erreurs par rapport aux versions préalables testées par les évaluateurs sourds en juillet 2026, mais ne les a pas éliminées.

Où Google dit que l’outil ne doit pas être utilisé
La sortie comporte une couche inhabituelle de gouvernance. Google DeepMind a réuni un comité consultatif, le Comité consultatif AI de la langue des signes, qui rassemble des organisations sourdes, notamment l’Association nationale des sourds, la Fédération mondiale des sourds, le Réseau professionnel des arts sourds et l’Institut national technique pour les sourds de l’Institut de technologie de Rochester. Le comité a co-rédigé un rapport d’impact conjoint qui définit ce que la technologie est et où elle s’arrête.

Le rapport définit SL2T 1.0 comme un outil d’aide à la génération de brouillons pour des paramètres informels à faible enjeu : messagerie, recherche, navigation, prise de notes et conversations de personne à personne occasionnelles. Il exclut explicitement les consultations médicales, les procédures judiciaires, les interactions avec la police, l’enseignement en classe, les entretiens d’embauche et les déterminations de prestations gouvernementales. Il indique également que l’outil ne satisfait pas aux obligations légales en matière d’adaptations raisonnables en vertu de la loi sur les Américains handicapés ou de cadres équivalents, et qu’il ne doit pas être utilisé par les institutions pour se substituer à des interprètes humains certifiés.

Le signeur reste dans la boucle tout au long du processus. Les deux applications affichent une prévisualisation du texte que l’utilisateur peut examiner et modifier avant de l’envoyer, et dans Live Transcribe, l’utilisateur sourd contrôle quand le texte traduit est affiché à un partenaire de conversation. Le rapport note que cette garantie suppose une littératie fonctionnelle en anglais pour détecter les erreurs.

Comment SL2T se comporte dans les limites du modèle
Le rapport d’impact recense les limites techniques du modèle en détail. La précision se dégrade dans les conditions de faible luminosité, de contre-jour intense ou lorsque les vêtements réduisent le contraste avec les mains et le visage. Le suiveur de pose ne suit que le sujet le plus grand dans le cadre et ne peut pas gérer plusieurs signeurs. Les performances chutent à des angles de caméra extrêmes ou lorsque le signeur est allongé. Les clips d’entrée sont limités à 60 secondes, et chaque clip est traduit de manière indépendante, sans mémoire des tours de conversation précédents. Le modèle n’a pas été formé ou évalué sur des signeurs de moins de 18 ans. Il ne prend en charge aucune liste de mots personnalisés, aucun nom de signe ni aucune préférence de dialecte.

Les mises à jour à court terme déjà sur la feuille de route incluent la dictée de signes de ponctuation, de nouvelles lignes, d’émoticônes et de commandes de base d’édition, ainsi que la mémoire de conversation sur des clips séquentiels dans Live Transcribe. Les objectifs de recherche à long terme incluent une meilleure sensibilité aux marqueurs non manuels tels que les expressions faciales et la position des sourcils, une prise en charge de plusieurs signeurs et une expansion de la collecte de données sur les dialectes régionaux de la langue des signes américaine et la langue des signes noire américaine.

Le projet a été lancé par Sam Sepah, un Googler sourd, et les perspectives sourdes ont été intégrées dans la collecte de données, les études d’utilisateurs et l’évaluation. Google décrit le lancement de SL2T comme le début d’un effort plus long : des langues des signes supplémentaires, la génération de la langue des signes et des capacités de frontière plus larges sont toutes répertoriées comme travaux actifs. La fonctionnalité est livrée sur Pixel 11 sans frais supplémentaires, avec davantage d’appareils à suivre, dans la collecte de données, les études d’utilisateurs et l’évaluation. Google décrit le lancement de SL2T comme le début d’un effort plus long : des langues des signes supplémentaires, la génération de la langue des signes et des capacités de frontière plus larges sont toutes répertoriées comme travaux actifs. La fonctionnalité est livrée sur Pixel 11 sans frais supplémentaires, avec davantage d’appareils à suivre."

https://www.unite.ai/fr/google-deepmind-brings-sign-language-translation-to-phones-with-sl2t/
#metaglossia
#metaglossia_mundus

La Défense nationale nie avoir retiré le statut «bilingue» à certaines unités

Il s’agit d’une question «d’uniformisation», dit le directeur des langues officielles du ministère de la Défense.

 

« Il n’y a personne à qui on interdit de parler le français. » Changer la désignation linguistique d’une centaine d’unités et organisations de la Défense au Canada a servi à « rectifier le tir d’un système désuet », affirme en entrevue le capitaine de vaisseau Vincent Pellerin, directeur des langues officielles du ministère de la Défense nationale.

 

Le Devoir révélait au début de la semaine dernière qu’une centaine de bureaux de la Défense nationale avait changé de désignation linguistique en raison d’une « nouvelle approche en matière de langue de travail ». Résultat ? Des employés ont perdu le droit de travailler en français dans leurs unités depuis janvier 2025.

 

La décision a été prise « pour augmenter la cohérence et la standardisation », affirme en entrevue M. Pellerin. C’est le ministère de la Défense qui a contacté Le Devoir après la publication de l’article, pour expliquer son raisonnement.

 

« On avait des unités dans la même région géographique, même dans la même base, qui avaient des langues différentes. Puis on avait beaucoup d’unités, en fait, qui n’avaient aucune désignation linguistique. Donc, on avait besoin de rectifier le tir sur ce système-là. On a adopté le système du gouvernement fédéral qui existe depuis 1977. »

 

Autrement dit, le ministère de la Défense s’est appuyé sur une disposition de la Loi sur les langues officielles lui permettant de désigner la langue de travail selon la région dans laquelle l’unité se trouve.

 

Vincent Pellerin

« Ce qui est important à souligner, c’est que ça n’impacte aucunement les services qui sont offerts aux Canadiens, aux Canadiennes, les services personnels des membres militaires et de leur famille. » Effectivement, la directive s’applique à l’interne, pour les employés du ministère.

 

Retirer la désignation « bilingue » de certains bureaux est-il une perte pour ces employés ? « Non. C’est que l’anglais est le minimum, pas que le français est interdit », affirme-t-il sans détour.

 

« On a pris l’approche que le gouvernement fédéral en entier utilise à travers le pays et on encourage l’utilisation des deux langues officielles, peu importe la désignation de langue de travail, parce qu’on reconnaît que le bilinguisme est une force et une richesse au sein du ministère de la Défense nationale. »

 

Même que, ajoute-t-il, « le bilinguisme est la clé pour être promu. Quand on regarde nos officiers aux grades supérieurs dans la Défense nationale, [on voit que] les francophones sont très, très bien représentés, parce qu’ils ont tendance à être plus bilingues ».

 

Vives réactions

Le syndicat de la majorité des employés de la Défense nationale s’est dit cette semaine « préoccupé » par la nouvelle, affirmant qu’il observera de près les effets de la directive sur ses membres.

 

Différents membres de la sphère politique ont aussi vivement réagi à la nouvelle. Le Parti québécois a publié ceci sur sa page Instagram : « Le Canada ne se donne même plus la peine de faire semblant d’être un pays officiellement bilingue. On s’en doutait, mais c’est maintenant officiel. Il y a maintenant deux réalités linguistiques au Canada. Le Canada anglais, où l’anglais est la seule langue officielle, puis le Québec et l’Acadie, où l’on a… l’anglais et la traduction officielle. »

 

Le député du Bloc québécois Mario Beaulieu a quant à lui affirmé sur sa page Facebook que le gouvernement enlève « les désignations bilingues pour que les plaintes qui augmentent ne soient plus considérées comme justifiées par la commissaire aux langues officielles »."

10 août

Mathilde Beaulieu-Lépine

https://www.ledevoir.com/politique/canada/1000802/defense-nationale-nie-avoir-retire-statut-bilingue-certaines-unites

#metaglossia 

#metaglossia_mundus 

"Barcelona rolls out a real-time AI-powered automatic translation system to improve citizen services

The pilot project will be introduced at Citizen Service Offices and Social Services to help overcome language barriers during initial administrative procedures.

 

05/08/2026 - 12:59 h

Barcelona City Council has launched a project based on portable devices that provide instant, two-way translation. The initiative, powered by artificial intelligence, has been designed to facilitate communication in situations where language barriers make understanding difficult. Its aim is to deliver faster, more effective and more accessible public services, promoting equal access to municipal services while strengthening social inclusion.

 

Greater autonomy for municipal staff

 

The initiative will enable municipal staff to hold immediate, two-way conversations in multiple languages, improving both accessibility and the quality of public services. It will also enhance the collection of information during initial interactions, reducing reliance on external interpreters at the early stages of communication and increasing staff autonomy when dealing with multilingual situations, ultimately streamlining administrative procedures.

 

Barcelona City Council emphasises that the device has been designed exclusively as a support tool for non-assessable administrative procedures and will always operate under the supervision of municipal staff. It will neither automate decision-making nor replace the role of public service professionals.

 

The rollout will begin in departments with the greatest exposure to multilingual environments. Within citizen services, the devices will be introduced at Citizen Service Offices and municipal information points, where they will be used to receive enquiries and collect basic information.

 

The technology will also be deployed across the direct intervention network of the Barcelona Municipal Institute of Social Services. This includes Social Services Centres, the Office for Social and Economic Benefits, the Social Guidance and Support Service for Homeless People, and other frontline teams operating throughout the city."

https://ajuntament.barcelona.cat/institut-innovacio-tecnologia/en/latest-news/news/barcelona-rolls-out-a-realtime-aipowered-automatic-translation-system-to-improve-citizen-services-1651839

#metaglossia 

#metaglossia_mundus 

Simpson World Guidebook Overcomes Rejection, Wins Fantastic Praise Three-year effort by 14 experts translates humor, adds cultural annotations for Korean fans

 

"21,000 sheets of manuscript paper, 1,200 pages. The guidebook *Simpson World* (Wilbook), containing all 20 seasons of the Emmy Award-winning animation *Simpson Family* (37 awards), weighs 4 kg. As American cartoonist Matt Groening, who created *The Simpsons*, put it, “It’s a dangerous brick book that could threaten your life if it falls on your head.” To overcome the local copyright holder’s threat, “If you can’t properly translate the unique humor, we’ll terminate the contract,” 14 people including translators, editors, and designers worked together for three years. Recently, we met them at Wilbook’s office in Mapo-gu, Seoul, to hear about the production process.

 

The book, which excited *Simpson* fans worldwide, has been long awaited by Korean fans. The domestic retail price is 170,000 Korean won. Imported copies of the original book exceed 300,000 Korean won even in the secondhand market. Editors said they saw one sold for 1 million Korean won at the used bookstore Aladin. This book, released locally in 2010, was attempted for publication by Wilbook in 2016 but was rejected by the copyright holder with, “Who do you think you are to dare publish *The Simpsons*? This is our work that’s impossible to translate.” Then, upon hearing “intelligence” that the copyright officer had been replaced three years ago, they began production negotiations.

 

The extensive content could not be handled by just one or two people. Translator Choi Se-hee said, “‘*Simpson Family*’ uses concepts ranging from American lower-class culture to modern philosophy, leaving it to the audience to figure out.” She added, “Even Americans often don’t fully understand, making the translation quite challenging.” When trying to reference the animated series’ translations broadcast domestically, there was a lot of information missing from the subtitles due to the physical limitations of TV screens. Thus, 14 people, including seven translators, had to step in. They created a group chat room to discuss and deliberate translation terms day and night. When expressing English-style humor in Korean, everyone had to pitch in and brainstorm. For example, when Bart, the eldest son of the Simpson family, receives a report card full of ‘D’ grades and is sarcastically told, “Have a D-lightful summer,” it was translated as, “D-fully enjoyable summer.”

 

Part of the contents of 'Simpsons World' /Park Seong-won

Editing and design were also challenging. Designer Kim Joo-yeon said, “We joked among ourselves that even if we reach our 60th birthdays, we might not finish.” The publisher instructed not to alter the format, photos, or graphics in any way. Typically, translating English to Korean increases the content by about 30%. They had to fit the information into the fixed layout, but to aid readers’ understanding, there was too much context of American culture to explain. Translator Shin So-hee said, “Thus, a special ‘annotated collection’ (*Simpson Universe Footnotes*) was created exclusively for the Korean edition.” The supplementary booklet, provided separately, includes background information on American history, politics, society, and culture necessary to understand *The Simpsons*. For example, the ‘Paul McCartney tombstone’ mentioned in the story refers to the conspiracy theory that McCartney died during the Beatles’ active period. After three years of work to finally complete it, the U.S. copyright holder had local Korean translators scrutinize every detail. Then came the message: “Fantastic!”

 

The main book and supplement (left) of 'Simpsons World' /Park Seong-won

The final challenge was sales. Although many are *Simpson* fans, there were worries about whether it would sell well. A crowdfunding campaign was launched, and the initial 1,000 copies sold out within 30 minutes. During the second round, 2,000 more copies were quickly exhausted. 419.83 million Korean won was raised instantly. It also became a hot topic at the recent Seoul International Book Fair. After completing the work, the production team found that *The Simpsons*’ stories appeared differently. Editor Park Hyo-joo said, “As a child, *The Simpsons* seemed like someone else’s world, and Marge, the mother, appeared frustratingly old-fashioned.” She added, “Now, I realize it overlaps with our own mothers and our lives, which feels new.”"

https://www.chosun.com/english/travel-food-en/2026/08/05/7LZ2JYS4PVBBNMKEHJNH2V6ABY/

#metaglossia 

#metaglossia_mundus 

"Deadline: 03-Sep-2026
The Danish Arts Foundation Translation Grants support Nordic publishers translating Danish literary works into other Nordic languages. The programme aims to strengthen the circulation of high-quality Danish literature across the Nordic region by funding professional translation and publication projects across a wide range of literary genres.

Eligible publishers must secure translation rights, sign a contract with a translator, and apply before the translation project is completed. The programme prioritizes artistic quality and supports professionally produced literary translations that expand access to Danish literature across Nordic countries.

About the Danish Arts Foundation Translation Grants
The Danish Arts Foundation invites applications from Nordic publishers seeking financial support to translate Danish literary works into other Nordic languages. The programme is designed to strengthen cultural exchange across the Nordic region by making Danish literature accessible to wider audiences through high-quality translations.

The initiative encourages professional publishing projects that preserve literary quality while increasing the visibility of Danish authors and literary traditions throughout the Nordic countries.

Programme Objectives
The programme aims to:

Support the translation of Danish literary works into Nordic languages.
Promote the publication and circulation of Danish literature across the Nordic region.
Strengthen cultural exchange between Nordic countries.
Increase access to high-quality literature for Nordic readers.
Support publishers producing professionally translated literary works.
Encourage literary diversity across fiction, non-fiction, children’s literature, comics, and essays.
Support special feature issues of magazines intended for Nordic audiences.
Ensure artistic quality remains primarily human-created, with technology used only in limited supporting or experimental ways.
Funding Purpose
The Translation Grants are intended to support publishers in covering the costs associated with translating and publishing Danish literary works for readers in other Nordic countries.

The programme helps publishers invest in professional literary translation while encouraging wider distribution of Danish literature across the Nordic cultural landscape.

Eligible Literary Genres
Funding may be awarded for translations in areas including:

Fiction for adults
Children’s literature
Young adult literature
Essays
General cultural literature
Literary non-fiction
Comics
Graphic novels
Special literary magazine issues intended for Nordic readers
Funding Benefits
The programme provides financial support for eligible translation and publication projects that:

Translate Danish literary works into another Nordic language.
Support professional literary translation.
Promote publication by Nordic publishers.
Increase cross-border literary exchange.
Expand access to Danish literature throughout the Nordic region.
Applicants should refer to the official programme guidelines for the exact funding amount available for each project.

Who is Eligible?
Applicants must meet the following eligibility requirements:

Be a Nordic publisher.
Intend to publish a literary work translated from Danish into another Nordic language.
Apply in the Nordic country where the original work was first published.
For Danish works, submit the application to the Danish Arts Foundation’s Committee for Literary Project Funding.
Secure the copyright or translation rights before applying.
Have a signed contract with the translator before submitting the application.
Submit the application before the translation project has been completed.
Who is Not Eligible?
The programme does not support:

Research publications.
Educational materials.
Regional or local history publications.
Handbooks.
Hobby books.
Self-help books.
Catalogues.
Translation of an applicant’s own works.
Text revision or proofreading projects.
Danish publishers seeking support for publishing Danish works in translation under this programme.
Completed projects submitted after the application deadline.
Eligible Expenses
Funding may support costs directly related to:

Professional literary translation.
Publication of translated literary works.
Translation projects for eligible literary genres.
Publication of Nordic literary magazine special issues.
Artistic Quality Requirements
Applications should demonstrate:

High literary and artistic quality.
Professional translation standards.
Strong publishing plans.
Cultural value for Nordic audiences.
The Danish Arts Foundation emphasizes that artistic quality should remain fundamentally human-created, while artificial intelligence or other technologies may only be used in limited supporting or experimental capacities.

Why This Programme Matters
The Translation Grants play an important role in strengthening Nordic cultural cooperation by:

Increasing access to Danish literature.
Supporting professional literary translators.
Encouraging publishers to invest in literary translation.
Promoting Nordic languages and shared cultural heritage.
Building stronger literary connections across Nordic countries.
How to Apply
Applicants should follow these steps:

Confirm that the publishing house is eligible.
Select an eligible Danish literary work.
Secure translation and publishing rights.
Sign a formal contract with the translator.
Prepare the publication proposal.
Apply through the appropriate Nordic funding authority.
For Danish works, submit the application to the Danish Arts Foundation’s Committee for Literary Project Funding.
Ensure the project has not been completed before the application deadline.
Tips for Applicants
Obtain translation rights before submitting the application.
Sign the translator agreement early.
Demonstrate the literary significance of the selected work.
Explain the publication strategy and target readership.
Ensure the translation project aligns with programme objectives.
Submit the application well before publication begins.
Common Mistakes to Avoid
Applying after the project has already been completed.
Failing to secure copyright permissions.
Applying without a signed translator contract.
Proposing ineligible publication categories.
Requesting funding for text revision instead of translation.
Submitting an application through the wrong national funding body.
Frequently Asked Questions (FAQs)
Can Danish publishers apply?
Generally, Danish publishers are not eligible to receive funding under this programme for publishing Danish works in translation.

Who can submit an application?
Eligible Nordic publishers planning to publish Danish literary works translated into another Nordic language.

Which literary genres are supported?
The programme supports fiction, children’s literature, young adult literature, essays, literary non-fiction, comics, graphic novels, and certain literary magazine issues.

Is a translator required before applying?
Yes. Applicants must have a signed contract with the translator before submitting the application.

Can completed projects receive funding?
No. Projects completed before the application deadline are not eligible.

Are self-help books or educational publications eligible?
No. Research publications, educational materials, self-help books, hobby books, handbooks, catalogues, and similar publications are excluded.

Where should publishers apply?
Applications for translations of Danish works should be submitted to the Danish Arts Foundation’s Committee for Literary Project Funding, while publishers in other Nordic countries should follow the relevant national submission requirements.

Conclusion
The Danish Arts Foundation Translation Grants provide valuable support for Nordic publishers committed to bringing Danish literature to readers across the Nordic region. By funding professional translation and publication projects, the programme strengthens literary exchange, promotes Nordic cultural cooperation, and increases access to outstanding Danish literary works across multiple languages."
https://www2.fundsforngos.org/individuals/nordic-translations-grants-for-publishers-denmark/amp/
#metaglossia
#metaglossia_mundus

"Considering the Many Forms a Family Can Take in Japanese Literature
Laurel Taylor on Maiko Seo’s Someone to Cook For
Via Europa Editions

Growing up, my mother and I often joked that I was a thirty-year-old in a ten-year-old’s skin. That happens when you’re an only child with a tendency toward bookishness—you spend all your time around adults, read much more than the average fourth-grader, and generally comport yourself with the same calm and poise you imagine adults have. For better or for worse, you spend a great deal of time making your own fun with an overactive imagination, taking care of yourself while your parents work, and picking up more responsibilities and self-recriminations than is probably healthy for a ten-year-old. Little do you know that adults, even the ones who seem put together, are just as messy as the children they’re raising. Or at least, this was my experience. This is why when I started translating Maiko Seo’s Someone to Cook For (Europa Editions, 2026), I recognized a kindred spirit in the protagonist, Yuko Morimiya. She too is an old soul in a teenager’s body, bearing a mantle of responsibilities and self-conscious habits like a well-worn and familiar coat.

My upbringing wasn’t quite as colorful as Yuko’s though. Unlike me, the poor girl had had five parents by the time she entered high school. Of course Yuko would balk at my calling her a “poor girl.” Throughout Seo’s novel, Yuko insists that her unusual childhood hasn’t been the tragedy everyone seems to expect—she’s loved all her parents and all her parents have done their best by her, misguided though they sometimes are. At the same time, however, Yuko is aware that because she’s moved households and changed names so often, there is an “acute sense of distance” between her and each new guardian. Where other teenagers are pulling away from their parents, chafing to become adults already, Yuko is constantly thinking about how she can cause her parents, new and old, the least amount of friction and worry.

[Yuko] is acutely aware of her own ever-shifting identity and how it is interwoven with the people who come in and out of the revolving door of her life.

Yuko’s sensitivity to her situation is partially built on the fact of blood relations, which remain a strong determining factor in family structures in Japan. As a toddler, Yuko loses her biological mother to a pedestrian hit-and-run incident, and her father, Shūhei Mito, remarries to a woman named Rika Tanaka when Yuko is in elementary school. However, at the age of ten, Shūhei announces that his company has transferred him to Brazil. Rather than follow him to the other side of the planet, Rika requests a divorce, and she and Shūhei put the question to Yuko: will you stay or will you go? For me, this is the moment Yuko goes from little girl to a little girl trying desperately to be an adult. Torn between her love for her father and her desire to remain in Japan, Yuko ultimately chooses Rika, and from that moment on, she never lives with a biological parent or relative again.

What this means legally speaking is that Yuko is removed from her father’s family registry (koseki) and placed on Rika’s—she transforms from Yuko Mita to Yuko Tanaka, and as a result, she is no longer a legal member of either of her biological parents’ family units. Until April 2026, Japan did not have any joint custody system in place; prior that change, children could not be listed on two family registries at once. Parents could, of course, negotiate visitation rights and child support in a more informal manner, but unlike many other wealthy countries, until 2026, these parental support systems were not obligatory; children could only ever legally be a member of one household, and the other household forfeited all rights and obligations to the child. For a ten-year-old like Yuko, however, the nitty gritty details of visitation and finances are less important than the very real fact of separation. It is evident in the novel that Shūhei loves Yuko very much, but his transfer to Brazil makes maintaining a strong relationship nearly impossible.

Over the course of a year, Yuko never hears from Shūhei at all and eventually gives up on their relationship. “We might have been father and daughter, but now we were apart—our time was over. My own life was right there in front of me, and I needed to focus on the people around me.” Yuko, old soul that she is, quickly learns that Rika’s ability to manage household finances is atrocious, so she takes it upon herself to make sure she and Rika have food and can make the rent on time. Though she is the child, Yuko very much grows into a parent.

Rika, through a series of remarriages and divorces, shuffles Yuko along from family registry to family registry, and with each new household, Yuko again calibrates herself to the adults around her, a process she slowly comes to realize is not the norm in other families. She is baffled by the way her friends Moe and Fumina rail against their parents, especially when Yuko is always holding back at home, forever walking on eggshells. When she and Sōsuke Morimiya, her third father, have a fight, she’s terrified of how to resolve the matter with him.

Maybe we should open up and talk about everything, explain to each other what we believed a father and daughter ought to be. The idea terrified me. I feared if we said everything we needed to say, we wouldn’t be able to stand each other anymore—neither of us actually knew what it was to be parent and child.

Though she would balk at my words, again I say “poor Yuko.” Three fathers and two mothers deep, she is only ever unsure about what a parent looks like or how they are “supposed” to act. And because Yuko is so self-sufficient and has grown up so much faster than other teenagers, she never reaches out for help, even to those who raised her before. She is convinced that to do so would be an unnecessary burden on her former parents, especially when they no longer have any obligations to her.

Though she tries her best to go it alone, ultimately she learns she can lean on the people who care for her and listen to what they have to say about her character and her choices in life.

In the process of editing this translation, one of my editors came back to me with questions about why Yuko is so reluctant to maintain contact with the parents who are no longer in her life. There is, of course, the legal argument; with each divorce, Yuko becomes part of a new family and sheds the old. More importantly, however, whether Yuko is conscious of the bureaucracy and paperwork behind all these family ties binding and unbinding, she is acutely aware of her own ever-shifting identity and how it is interwoven with the people who come in and out of the revolving door of her life. Better to go it alone than reach out for someone who might be gone in a year, right Yuko? Again, I sense a kindred spirit in her; as my friends and family can attest, I also have a hard time asking for help, even from those who would happily give it. Luckily for both Yuko and I, very often help comes even when neither of us seek it.

In translating Someone to Cook For, I can’t add an entire paragraph or even a set of little notes here and there explaining the differences between Japan’s divorce system when compared to those in the UK or the US. What I can do is concentrate on my empathy with Yuko, on the ways in which her thought processes mirror my own. I finished translating the first draft of this novel in June of 2025 and immediately burst into tears. My father had passed unexpectedly only a month prior, and the final scene of the book had touched that part of me that was still in mourning disbelief. Rika also reminded me a lot of him—a little too focused on shopping over family finances, caring but impulsive. Unlike Yuko, my parents never divorced, but still I felt her same uncertainty when faced with my new fatherless place in the world, and I could bring that uncertainty to bear when I began editing and refining the translation.

Who are we when our parents are no longer there to serve as a guidepost for us? I think the common American mythos would have us go on a journey of self-discovery; we would strike out alone, traveling or starting a new business, trying to discover ourselves via “bootstraps” style independence. In Yuko’s story however, she discovers herself not through stubborn independence, but instead through locating new family, new parental figures and mentors. Though she tries her best to go it alone, ultimately she learns she can lean on the people who care for her and listen to what they have to say about her character and her choices in life. I only hope I can be as lucky when it comes to finding new family in my future.

______________________________

Someone to Cook For by Maiko Seo, translated by Laurel Taylor, is available from Europa Editions.

Europa Editions family Japan Japanese Literature Laurel Taylor literature in translation Maiko Seo Someone to Cook For translated literature translation

Laurel Taylor
Laurel Taylor teaches Japanese at the University of Denver. She holds a PhD in the joint Japanese Literature-Comparative Literature track from Washington University, St. Louis, and an MFA in Literary Translation from the University of Iowa, where she was a recipient of the Iowa Arts Fellowship. Her translation work has appeared in Asia Literary Review and English PEN Presents. She currently at work on the translation of Maiko Seo’s All The Long Nights forthcoming from Europa Editions in 2027."
Laurel Taylor
August 6, 2026
https://lithub.com/considering-the-many-forms-a-family-can-take-in-japanese-literature/
#metaglossia
#metaglossia_mundus

"   Local large language AI models are closing the gap with cloud-based models in professional computer-assisted language translation, according to a new study from the University of Georgia that also sheds light on how large language models (LLMs) learn to translate.

The findings could give professional translators a practical alternative to cloud-based AI, particularly in fields where sensitive documents must remain offline and secure.

“There is a niche area of the translation market where translation must be done offline,” said Yuri Balashov, a professor in the Franklin College of Arts and Sciences Department of Philosophy and a faculty fellow in the Institute for Artificial Intelligence, who led the project. “In security, military, and patent-related translation projects, you need to shut down your internet. You cannot work online. But you can still run local models, so we wanted to explore how well they perform.”

The research team, including three graduate students who took Balashov’s seminar on translation technologies, compared nine locally run AI language models with leading commercial machine translation systems. To test the systems, they used a set of multilingual medical translations built from professionally translated versions of the Christopher and Dana Reeve Foundation’s paralysis resource guide.

“It was translated into several languages, including Japanese, Chinese, Russian, German, French, Brazilian Portuguese, and Hindi,” Balashov said. “With the foundation’s permission, we aligned some of them with the English original, sentence by sentence, and that is how we created our corpus.”

The study found that the strongest local models, including Translate Gemma and Mistral, performed competitively in translation quality with larger cloud-based language models such as ChatGPT and Claude, although leading commercial machine translation engines (DeepL for English to German, Russian, and Japanese; Baidu for English to Chinese) still produced the highest-quality translations overall.

“They are still behind, but the gap is narrowing and the field is changing very rapidly,” Balashov said.

The bigger question
Why can large language models translate so well despite never being explicitly trained as translators? Balashov argues that the answer lies in how the models learn. Rather than relying on a single source of information, they develop translation abilities through two concurrent methods: by studying pairs of matched sentences in two languages, and by reading large collections of text in each language separately. Gradually, he suggests, the model builds connections between languages despite not being taught explicit translation rules.

“Both processes are integral to the development of translation abilities of LLMs,” Balashov said. “They both contribute to it but maybe to different extents.”

Rethinking translation
To reach his conclusions, Balashov referred to one of machine translation’s oldest ideas: Interlingua, which holds that translation is not just about converting words from one language into another but capturing meaning behind phrases and sentences, before expressing that same idea in a different language. Once that meaning is encoded in a language-neutral form, it can be expressed in any language without losing the message.

Some researchers believe today’s LLMs may have revived this idea, which dates to early machine translation systems in the 1950s and 60s, through internal language-independent representations. But Balashov remains skeptical.

Translation is more flexible and context-dependent than Interlingua theories suggest, he said. There is no doubt that language models learn cross-lingual representations from exposure to multilingual data during training and fine-tuning. However, they remain dependent on the particular mix of languages used in training. Real-life translation is messy, opportunistic, and much less systematic than is assumed in any interlingual account.

“Your brain almost immediately outputs an equivalent of a shorter phrase in a target language,” Balashov said. “But there’s also another way to translate for us humans. We often slow down and reflect on a more complicated sentence. We try to understand what it really says in the source language, and then we try to transfer that meaning to a target language.”

For Balashov, that same interplay between fast pattern recognition and slower, context-driven reasoning may help explain not only how humans translate but also how LLMs acquire translation abilities.

Humans in the loop
The research is intended to help professional translators and smaller translation agencies adapt to a rapidly changing profession.

“Our target audience in a recent series of papers is human translators and smaller translation agencies, not high-tech labs like Google or Anthropic,” he said. “We think our audience could benefit from what I call translation analytics.”

Balashov describes translation analytics as a combination of “automatic evaluation of translation quality, human evaluation and lightweight statistical analysis” used to assess translation performance.

While AI continues to reshape the translation profession, he added, the role of human translators is evolving rather than disappearing.

“I would like to think that there will always be a need for professional translators. There are areas where the proverbial human expert in the loop is sorely needed,” Balashov said. “But the nature of the work that translators need to perform is changing very rapidly. Those who want to continue working in this area need to adapt very fast.”"
By Moses U. Okocha
https://research.uga.edu/news/how-can-ai-translate-without-being-trained-as-a-translator/
#metaglossia
#metaglossia_mundus

I'm 94 and still work five days a week. Six decades in, I still have no plans to retire
Marta Barany has been an interpreter and translator for over 60 years. There's a reason she keeps working

"I'm now 94 years old, interpreting and translating in both the private and public sectors (including Centrelink), and have no plans to retire soon.

A natural path
Becoming an interpreter was always a natural progression for me; I speak a few languages, and I love dealing with people.

I was born and raised for the first 12 years of my life in Hungary. But in 1945, near the end of World War Two, my parents and I fled as refugees to Germany after the Soviet Red Army invaded and occupied our country.

I already knew Hungarian, but I learnt German, English, Latin and French through my six years of schooling in Germany, which have been very helpful to know.

I arrived in Melbourne with my family in January 1951, after 65 days of travel, as displaced persons from the war.

I was 18 years old and still remember feeling a heat I had never experienced in my life when we were taken to Bonegilla, an ex-army camp near Albury in northern Victoria that was then used as a camp to process post-war European migrants.

Because the camp was full, we were put in an army tent — and that is how I started my life in Australia.

Finding my purpose
It wasn't until the late 1950s, when I was in my late 20s, a mother and working an admin role, that my interpreting career began.

The 1956 Hungarian Revolution brought a wave of Hungarian migrants to Australia in the years that followed. But there was no community interpreter service here at the time, and people needed help.

So I stepped forward as an interpreter. I believe it's important to be understood as a person — to say what you want to say and be able to express yourself. And if you don't speak the language of a place, that's very hard to do.

One memory that stands out in my career was working with the Department of Immigration's telephone interpreter service, when a phone call from a hospital somewhere in rural Western Australia came through late one evening.

A young woman was in labour, and I had to interpret for her and the medical team: telling her when to push, when not to push and that she was going to get through it.

Then, when she had given birth, I had to tell her husband he was a father to a "lovely baby boy".

It was a special moment, and I felt privileged that my work let me be a part of it.

As a 'clog wog', Australians didn't know I was a migrant, so they confided their racism
Across my six-decade-long career, I’ve also been involved in starting several of Australia's interpreting and translating services.

I was one of the founding members of the Telephone Community Interpreting Service and Australian Institute of Interpreters and Translators, and I helped set up the National Accreditation Authority for Translators and Interpreters — as well as the Hungarian branch of Catholic aid organisation Caritas.

It's been a privilege to help others feel understood, and I especially feel connected with the Hungarian communities both here and abroad.

Some Hungarian Australians have somewhat adopted me as their mother, which is a role I'm happy to fulfil.

I also chaired the Hungarian Olympic Committee during the 2000 Sydney Games and served as the Ethnic Affairs Commissioner of NSW. And I have been awarded the British Empire Medal, Order of Australia Medal, and this year, the Interpreters and Translators Medal.

If my mind works, I still work
At 94, I'm still working five days a week — usually three or four hours a day. I don't have any plans to stop working, but I know it'll be time to do so when my mind gives up.

I try to work as much as I can and feel energised by it.

If my mind works, I will still work.

I used to travel a lot for work as I mostly interpreted onsite. I worked in Cairns, Townsville and Brisbane in Queensland; in the courts in Adelaide; in Broken Hill and Lightning Ridge in rural NSW — all over the place.

But these days, I mostly work over the phone. I still drive, so I work onsite if I can find a parking spot — otherwise I do a lot of telephone and video call interpreting.

Although German was like a mother tongue to me, it's become rusty; so, I now only interpret in Hungarian and English Hungarian.

I get a pension and my freelance work does mean extra income — not as much as when I worked full-time, but it's still very good.

And my work keeps me busy.

In all truth, I wouldn't know what to do with my mind if I stopped working.

And I know that I'm good at my job and I'm needed."
5 August 2026
Marta Barany

https://www.sbs.com.au/news/insight/article/im-94-and-still-work-five-days-a-week-six-decades-in-i-still-have-no-plans-to-retire/s3e0ptjcb
#metaglossia
#metaglossia_mundus

Rights Complaint Filed Over Lack of Interpreters for Foreign Workers at Labor Board

"A complaint has been filed alleging that regional and central labor relations commissions discriminated against a foreign worker's right to equality by conducting hearings without providing interpretation, in a case where the worker had sought relief for unfair dismissal.

Korean Lawyers for Public Interest and Human Rights, known as Duroon, said on the 4th that it submitted a complaint to the National Human Rights Commission of Korea on the 3rd of this month. The complaint requested a recommendation that "the Labor Relations Commission improve its practices and systems to provide interpretation when investigating, deliberating on, and questioning foreigners, and to take active measures such as guaranteeing the participation of a person in a relationship of trust and extending the duration of hearing sessions."

According to Duroon, two Sri Lankan foreign workers who had worked as welders at a shipyard in Ulsan filed for relief from unfair dismissal with the Ulsan Regional Labor Relations Commission in June last year after being dismissed from their jobs.

However, Duroon explained that the Ulsan commission did not provide interpretation to the workers, who were not proficient in Korean, at a hearing session held in August of the same year. As a result, a fellow worker with no professional interpreting experience served as the interpreter, making it difficult to adequately represent their positions. "The commissioners spoke in long and complex sentences without considering interpretation, used legal terms without separate explanation, and often continued speaking before interpretation was completed," Duroon said.

Duroon also claimed that it requested an extension of the hearing session time in consideration of the time required for consecutive interpretation, but the request was not accepted. As a result, in a hearing session that lasted about one hour and 20 minutes, actual questions and answers accounted for less than half of the time. According to Duroon, interpretation was likewise not provided and the hearing time was not extended during the National Labor Relations Commission's review procedure conducted in February this year.

Duroon said, "Because the Labor Relations Commission's unfair dismissal relief procedure is a preliminary procedure to administrative litigation and effectively has the character of a trial, interpretation support and time extensions that take this into account are necessary to guarantee foreign parties a substantive right to participate, just as in criminal and civil procedures.""
https://en.sedaily.com/society/2026/08/04/rights-complaint-filed-over-lack-of-interpreters-for
#metaglossia
#metaglossia_mundus

Amberscript associe la transcription par IA à une couche de contrôle humain

"La transcription et l’analyse de textes sont des tâches courantes dans divers secteurs. Pensez aux interviews pour les journalistes, aux comptes rendus de consultations médicales ou aux documents juridiques. Les outils de transcription vous offrent une solution en un clic pour ce processus chronophage.

Nous testons des alternatives européennes aux divers outils de transcription indispensables à notre rédaction. Notre oreille attentive a déjà analysé le Good Tape danois et le Gladia français ; c’est maintenant au tour du néerlandais Amberscript.

Langue maternelle
Amberscript a été fondée en décembre 2017 à Amsterdam, avec pour premier produit une fonction de reconnaissance vocale en néerlandais. En 2018, un service de transcription et de sous-titrage manuel a suivi. Entre-temps, l’entreprise propose plus de 90 langues différentes, des transcriptions automatiques et professionnelles, ainsi que du sous-titrage.

Un premier point positif : l’entreprise est basée aux Pays-Bas. En tant que journalistes, nous traitons régulièrement des transcriptions en néerlandais. Nous nous attendons donc à ce que la qualité de l’outil néerlandais soit élevée, surtout par rapport à Gladia qui a obtenu un score légèrement inférieur pour notre langue maternelle.

Lors d’un premier test avec Amberscript, nous remarquons que la transcription en néerlandais est très précise. Sur le test linguistique, l’outil obtient d’ores et déjà un dix sur dix.

Automatique versus professionnel
Amberscript propose deux options de transcription. Vous pouvez faire transcrire des fichiers audio automatiquement ou faire appel à un professionnel humain. Pour notre domaine d’application, à savoir la transcription d’interviews, la transcription automatique est suffisante.

Simple et direct
L’outil est simple et direct. Vous cliquez sur « Télécharger un fichier » et choisissez ensuite parmi différents formats de fichiers, tels que mp3, mp4, wav, m4a, m4v, mov et bien d’autres encore. C’est un avantage important : les transcriptions sont souvent enregistrées via Microsoft Teams ou un enregistreur, ce qui produit un format mp3 ou mp4.

Amberscript prend en charge différents formats de fichiers. Source : Amberscript
Dès que le fichier est chargé, vous recevez une notification par e-mail indiquant qu’il est prêt dans l’éditeur. Ce tableau de bord est également très clair. Dans le menu supérieur, vous pouvez modifier l’heure de début, la durée ou la vitesse. Vous pouvez également avancer ou reculer dans la transcription. De plus, vous avez la possibilité d’attribuer automatiquement des locuteurs ou d’en ajouter de nouveaux.

Dans l’éditeur, vous pouvez écouter et modifier les transcriptions. Source : Amberscript
En outre, vous pouvez également marquer des passages importants dans le texte, ou modifier ou remplacer des mots dans l’ensemble du document. Un petit détail qui nous manque ici est la possibilité d’ajouter des notes ou des commentaires aux passages marqués.

Traduire
Par ailleurs, vous pouvez faire traduire vos textes au sein du tableau de bord. Une fois que vous avez sélectionné une traduction, un menu déroulant apparaît, vous permettant de basculer facilement entre les langues. Cela peut être utile, par exemple, lorsque vous enregistrez une conversation de réunion en anglais, mais que la rédaction du compte rendu se fait en néerlandais. Vous pouvez ainsi vérifier certaines déclarations en anglais.

Prix par heure
Pour notre période d’essai, nous avons utilisé les dix minutes de transcription gratuites. Si vous voulez vraiment utiliser l’outil, vous devrez souscrire un abonnement. L’abonnement est à chaque fois basé sur le nombre d’heures de transcription. Le forfait Starter revient à 19 euros par mois et comprend 300 minutes. L’abonnement Pro offre 600 minutes pour 29 euros par mois, et l’abonnement Power (49 euros/mois) comprend 1 500 minutes de transcription.

Pour vous donner une idée : l’abonnement Starter revient à 3,8 euros par heure. Il comprend des fonctionnalités telles que plus de 90 langues, l’identification automatique des locuteurs et des options d’édition. Mais aussi :

Exportation des transcriptions en Word, JSON, TXT, CSV
Exportation des sous-titres en SRT, VTT, EBU-STL et plus
Certifié ISO27001 & ISO9001
Traitement et stockage des données sur des serveurs de l’UE conformément au RGPD
Le prix de l’abonnement d’entrée de gamme (Starter) est légèrement plus élevé que celui d’autres outils que nous avons examinés. Ainsi, le montant mensuel de l’abonnement le moins cher chez Notta s’élève à 11,69 euros, et à 16 euros pour Good Tape.

L’excellence en néerlandais
Outre la qualité de la transcription, nous visons également une approche européenne. Amberscript promet un traitement sécurisé des données. Ainsi, la plateforme est conforme au RGPD, certifiée ISO27001 et ISO9001, et possède le label TPN pour une excellente sécurité des données.

Cet outil de transcription amstellodamois possède toutes les fonctions nécessaires pour transcrire et traiter des interviews. Les abonnements sont plutôt onéreux, mais vous bénéficiez en échange d’une qualité locale.

Pro's
Transcriptions précises (néerlandais)
Convivial & clair
Traduction en temps réel
Contra's
Un peu plus cher
Notes impossibles
ITdaily teste des alternatives européennes et open source aux outils logiciels américains populaires que nous utilisons quotidiennement. Vous souhaitez découvrir plus de logiciels européens ? Grâce à cet aperçu, vous retrouverez tous les avis."
https://itdaily.fr/tests/logiciel/avis-sur-amberscript-transcription-impeccable-dans-les-frontieres-de-lue/
#metaglossia
#metaglossia_mundus

"La cérémonie de montée des couleurs organisée le lundi 3 août à la Cour suprême, à Porto-Novo, a été marquée par l’exécution de l’hymne national en langue dendi. Cette interprétation a été assurée par la chanteuse Amy-Mako, déjà remarquée pour la même prestation lors du défilé du 66ᵉ anniversaire de l’indépendance à Cotonou, le 1er août.

Selon les informations communiquées par le quotidien de service public La Nation, cette initiative prolonge une pratique instaurée depuis plusieurs années consistant à faire interpréter « L’Aube Nouvelle » dans les langues nationales à chaque premier lundi du mois.

Une nouvelle langue à l’honneur
À l’origine de cette initiative, le président de la Cour suprême, Victor Dassi Adossou, qui a souhaité inviter Amy-Mako après sa prestation lors des festivités de l’indépendance. Jusqu’à présent, l’hymne était régulièrement chanté en fon au sein de l’institution, tandis que l’apprentissage d’autres langues nationales se poursuit.

Le magistrat a expliqué que cette démarche vise à encourager une plus grande place des langues béninoises dans les cérémonies officielles. Selon lui, il est souhaitable de « parler nos langues nationales » et de « chanter en langue nationale », afin de renforcer leur usage dans la vie publique.

L’indépendance associée à l’identité nationale
Profitant de cette cérémonie, Victor Dassi Adossou est revenu sur la portée de la célébration du 1er août 1960. Il a rappelé que l’indépendance ne se limite pas à un anniversaire institutionnel, mais renvoie aussi à la capacité des citoyens à construire leur avenir en s’appuyant sur leurs propres ressources.

Selon le président de la Cour suprême, « être indépendant, c’est penser par nous-mêmes ». Il a également évoqué la nécessité de promouvoir les productions nationales, les langues locales ainsi que les expressions culturelles béninoises.

Un rappel du rôle de la justice
Le discours a également porté sur les missions de la justice dans le fonctionnement de l’État. D’après Victor Dassi Adossou, le choix du Bénin en faveur de la démocratie pluraliste depuis la Conférence nationale des forces vives repose sur l’existence d’un État de droit soutenu par une justice indépendante et efficace.

Le président de la Cour suprême a estimé que les magistrats contribuent à la cohésion nationale en rendant des décisions conformes au droit. À ses yeux, « travailler pour le bien collectif » constitue l’une des traductions concrètes de l’indépendance.

Le dendi rejoint ainsi les langues déjà utilisées lors de cette cérémonie mensuelle. Le Bénin compte plus d’une cinquantaine de langues et variétés linguistiques parlées sur son territoire. Depuis plusieurs années, les pouvoirs publics et certaines institutions multiplient les initiatives destinées à promouvoir ce patrimoine linguistique dans les espaces officiels, aux côtés du français, langue officielle du pays."
04/08/2026
Armel TOGNON
https://lanouvelletribune.info/2026/08/benin-lhymne-national-chante-en-dendi-a-la-cour-supreme-apres-le-1er-aout/
#metaglossia
#metaglossia_mundus

"In Praise of the Translators Who Transport Us Across the Globe
In Feature Articles by Olivia SnaijeJuly 29, 20262 Comments

Literary translations are investments in cultural infrastructure, a public good that can be applied to all publishing markets, no matter the language territory.

By Olivia Snaije, Contributing Editor

Much ink has been spilled this summer about Christopher Nolan’s blockbuster based on Homer’s The Odyssey. Nolan has said that he was inspired by Emily Wilson’s 2017 novel translation, the first into English by a woman, and one that is resolutely modern. (Wilson is halfway through a re-translation of The Odyssey in which she says she won’t confine herself to the same number of lines as the original, even if it will still be in iambic pentameter.)
Since the release of Nolan’s film, Wilson’s translation of The Odyssey shot to No. 3 on Amazon’s bestseller list and No. 11 on Barnes & Noble’s, and sales in the U.S. have reportedly surpassed one million copies. This is a rare occurrence for a translator. Before publication, it took the classicist five years and considerable prowess to translate the work. She has said she never imagined such success.

‘Unsung Heroes’
Translators are like sculptors, chipping away at a block of stone, choosing one particular word out of several, immersed in the source language and dialogues which will certainly reflect a dialect or the country among many possible countries. They will draw on political, historical, social, cultural and even personal considerations, while dealing with grammatical quirks, layered, ambiguous plots, and distinct writing styles. They may even experience secondhand trauma when the subject of a book is particularly grim.

Then consider the practical hurdles: not only do translators often act as scouts and agents combined (sometimes pitching a book for years), but they deal with long working hours and poorly paid commissions.

Meanwhile, there has been much justified concern about the impact of AI on translation. Following a survey in 2024 by the U.K. Translators Association (part of the Society of Authors) a number of translators reported losing work to generative AI, and 50% said that AI had decreased their income. Three-quarters of translators said that AI would negatively impact their future income.

The Society of Authors has called for industry solidarity with human translators. After all, once a book in translation has been published, the reader is magically transported to another country, and sometimes to another world. But, behind the magic, are years of skill, sacrifice, and often brilliance.

Translators are, and have traditionally been, the unsung heroes who have remained in the shadows—even if many changed the world throughout history. In recent years, they have begun to emerge from the shadows to some extent—awards such as the International Booker Prize and the Dublin Literary Award have done much for translators’ exposure. But it’s not nearly enough.

‘The Vanguard of Literary Change’
Perhaps what our story-loving societies need are more stories about translators themselves, and their craft. In the July issue of the new magazine Now Voyager, for example, a longform article by Makana Eyre takes a deep dive into the woes of a young literary translator and a contemporary Russian novel. Eyre tells the story of Bela Shayevich, who hadn’t yet translated a major novel but was commissioned by an editor at Farrar, Straus and Giroux (FSG), who felt she had all the necessary skills to translate Ludmila Ulitskaya’s 2010 novel, The Big Green Tent.

Bela Shayevich (credit: University of Iowa).

Ulitskaya’s literary agents, Yulia Dobrovolskaya and Alexander Klimin at Elkost were skeptical. They wanted a better-known translator. But Shayevich was backed by Ulitskaya’s editor. Then, he left FSG and the book was assigned to a different editor. Shayevich finished the translation under tremendous pressure, and an excerpt was even published in The New Yorker in May 2014.

But one of Ulitskaya’s agents, Klimin, continued to have misgivings due, in part to Shayevich’s more contemporary approach to the translation. Ulitskaya was apparently happy with Shayevich’s translation.

Eyre quotes Shayevich on her approach to translation:

“‘If I say bread, depending on where you’re from, you’re going to see something really different,’ she told me, alluding to a passage in Walter Benjamin’s essay “The Task of the Translator.” A Russian might visualize a loaf of black bread, a French person might see a baguette, an American might see Wonder Bread. ‘Even calling bread bread is a mistranslation,’ Shayevich said.”

In the end, FSG commissioned the more experienced Polly Gannon, a professor with a Ph.D. in Russian literature, to translate a second version of The Big Green Tent which was published in November 2015. Shayevich’s translation only exists in the form of review copies sent out beforehand.

Shayevich, however, was eventually vindicated in 2018 when she won the inaugural TA First Translation Prize for her translation of Svetlana Alexievich’s Secondhand Time. Jacques Testard of Fitzcarraldo Editions had commissioned Shayevich in 2015 for the translation (before Alexievich won the Nobel Prize in Literature later that year) and when Secondhand Time was published in 2016 her translation was highly praised.

Daniel Hahn (credit: Camila França Photography).

The TA First Translation Prize was established in 2017 with help from the award-winning translator Daniel Hahn, who donated half of his winnings from the Dublin Literary Award from that same year for his translation from Portuguese of José Eduardo Agualusa’s A General Theory of Oblivion.

Hahn wrote about setting up the prize in an article in The Guardian. His title was “Translators are the vanguard of literary change: we need better recognition.”

In between his translations, Daniel Hahn published a book this past spring which is an investigation of how Shakespeare’s plays are translated for global audiences, and the skillfull wordplay used by translators across the world. It involved extensive interviews with approximately 14 translators working in languages from Hungarian and Danish or Swahili to Tswana, Māori, and Bangla.

If This Be Magic: The Unlikely Art of Shakespeare in Translation, is an ode to his fellow translators. In a recent interview with Anil Sarvepalli, the host of Harshaneeyam, one of the most dynamic and well-researched podcasts about literary translation today, Hahn said that when translating a writer as good as Shakespeare, nothing in his writing is accidental. So when translating Shakespeare—or any writer whose work is highly kinetic—you want to keep everything.

But keeping everything requires keeping all the properties of the writing, which poses several challenges. “Every language works differently, and you have to notice everything in the first place,” said Hahn. In the end, “you need to preserve everything to keep Shakespeare as Shakespeare,” he explained. “But as all translators know, keeping everything also means changing everything.”

For Lin King, the translator of Yáng Shuāng-zǐ’s, Taiwan Travelogue which won this year’s International Booker Prize, it was her first time translating a full-length novel, a project that grew out of her masters thesis.

Author Yáng Shuāng-zǐ, and translator Lin King, during the International Booker Prize 2026 ceremony (credit: Booker Prize Foundation).

King was also interviewed on Harshaneeyam before the book won the prize, as well as on another podcast about translation, It’s Lit in Translation, run by translator and bookseller Evelyn Heis. In both interviews King describes the multi-layered opacity of the book, which is a fictional account disguised as a translation of a rediscovered work by a Japanese author. Thus, the translation required knowing not only Mandarin and Hokkien (spoken in Taiwain) but Japanese and archaic Japanese terms.

On the contractual side, Lin told Heis that Taiwan Travelogue received several offers from publishers in the U.K, but one of the reasons Shuāng-zǐ and King chose to go with And Other Stories was because, like their U.S. publisher, Graywolf, they put translators names on the cover of their books, which many publishers still do not do. That kind of support—acknowledging the role of a translator with cover credit, is an essential recognition.

An Investment in ‘Cultural Infrastructure’
As a 2026 ThinkPub study on Books in Translation in the European Publishing Market stressed, literary translations are more than commercial transactions, they are investments in cultural infrastructure.

“Translation must be understood both as a market activity and as a public good,” the ThinkPub report concluded. “Commercial logic alone is not sufficient to sustain linguistic diversity, particularly for works from smaller languages or less commercially dominant regions. Public support mechanisms—including translation grants, publisher subsidies, and international cooperation programmes—emerge as decisive factors in maintaining diversity and ensuring long-term cultural circulation. Indeed, the idea that translation is a public good can applied to all publishing markets, no matter the language territory.

With fall book fair season right around the corner, the LitAg Center at the Frankfurter Buchmesse will soon be buzzing with agents buying and selling translation rights. Which makes now the perfect time to recognize—and praise—the vital, multidimensional work of literary translators and their enduring cultural contributions. And more to the point, how we can better support them.

Olivia Snaije

https://publishingperspectives.com/2026/07/in-praise-of-the-translators-who-transport-us-across-the-globe/
#metaglossia
#metaglossia_mundus

"New York : la ville aux 700 langues

New York est la ville où l’on parle le plus de langues sur la planète. Cette nouvelle Babel parviendra-t-elle à sauver les dialectes les plus rares du monde ?

 

Apportant avec eux les langues de leur pays natal, les immigrants débarqués par bateau sur Ellis Island attendent les contrôles officiels et l’approbation des autorités pour atteindre leur destination : la ville de New York, qu’ils aperçoivent déjà.

 

Le seke est une langue peu usitée que l’on parlait à l’origine dans cinq villages du nord du Népal. Mais son avenir pourrait dépendre d’une poignée de villages d'un nouveau genre : des immeubles du centre de Brooklyn, quartier de la ville de New York.

 

Comment cette langue, peu documentée et seulement parlée par moins de 700 personnes dans les sommets de l’Himalaya, en vient à être entendue dans cette jungle de béton ? Rasmina Gurung, âgée d'une vingtaine d'années, est l’une des plus jeunes personnes à parler le seke. Sa grand-mère le lui a appris, dans son village, mais elle a déménagé tôt dans la capitale, Katmandou, pour finalement arriver à New York. Elle estime qu’un quart de son peuple habite la ville. Ils rejoignent ici des locuteurs de dizaines d’autres langues en danger de tout l’Himalaya. Tous forment de nouvelles communautés où ils communiquent dans un mélange de népalais, de tibétain, d’anglais et de leur propre langue maternelle en danger.

 

Mais New York, la ville où l’on parle le plus de langues dans l’histoire du monde, pourrait avoir atteint son pic de diversité. Les plus de 700 langues qui y sont parlées représentent plus de 10 % de toutes celles utilisées dans le monde. Bien qu’elles soient en grande partie invisibles, et inaudibles, aux étrangers, les langues parlées à New York proviennent des quatre coins du monde. De nombreux immigrants sont arrivés de points névralgiques linguistiques au cours des dernières décennies. Des zones comme l’Himalaya, l’Afrique de l’Ouest, l’Asie du Sud-Est insulaire, ou encore d’Amérique Latine. Cependant, les forces qui ont réuni ces personnes commencent aujourd’hui à les séparer.

 

À travers ses cinq quartiers, la riche diversité linguistique de New York est impossible à ne pas voir dans ses rues, boutiques et arts de rue. D’annonces publicitaires en arabe ou en hindi, à des fresques murales où le mot « liberté » est écrit en bengali, chinois et espagnol.

 

À travers ses cinq quartiers, la riche diversité linguistique de New York est impossible à ne pas voir dans ses rues, boutiques et arts de rue. D’annonces publicitaires en arabe ou en hindi, à des fresques murales où le mot « liberté » est écrit en bengal, chinois et espagnol.

 

Au vu de l’accélération de la perte des langues, même au sein de leur contrée natale, des menaces qui pèsent sur l’immigration et de la hausse des coûts d’une vie en ville, le temps pourrait venir à manquer.

 

La remarquable convergence des langues à New York et dans les villes similaires pourrait rapidement disparaître, avant même que l’on ait le temps de la documenter ou de la soutenir. L’urgence est le moteur de l’Alliance des langues en danger, organisation que je co-dirige, qui a commencé à cartographier cet environnement linguistique.

 

Un ensemble sans précédent d’opportunités culturelles, scientifiques, éducationnelles et même économiques, est en jeu. C’est la première fois que les linguistes et les locuteurs sont si bien positionnés pour documenter les langues pour lesquelles peu de trace existe, si elles existent. Les possibilités artistiques, musicales et culinaires sont tout aussi exceptionnelles car des points de vue de toute la planète se mêlent dans un même espace.

 

Irwin Sanchez, chef cuisinier et poète du Queens, qui parle le nahuatl, auparavant la langue des Aztèques, cuisine des tacos, des moles et des tamales en gardant à l’esprit les significations originales de ces mots. Husniya Khujamyorova, qui parle le wakhi du Tadjikistan, écrit les tout premiers livres pour enfants à destination des locuteurs de six langues pamiries, aujourd’hui toutes représentées le long de la propre route de la Soie de Brooklyn. Ibrahima Traore, qui a fait tout le chemin depuis la Guinée jusqu’au Lower East Side, enseigne le n’ko, un système pionnier d’écriture d’Afrique de l’Ouest, et encourage son usage dans toutes les nouvelles technologies. Boris Sandler, locuteur yiddish et écrivain né en Moldavie, contribue à sa façon, roman après roman, à la renaissance miraculeuse du yiddish à New York.

 

Le lenape, la langue d’origine du lieu où s’est élevée New York, renaît également de ses cendres, contre toute attente. Depuis sa dernière place forte, dans la campagne de l’Ontario, où seul un seul locuteur natif subsiste, une nouvelle génération d’activistes présente la langue à une plus large audience. L’une d’elle était Karen Mosko qui, avant sa mort, se rendait à Manhattan, « l’île aux nombreuses collines » en lenape, une fois par mois pour enseigner la langue.

 

Et puis il y a Rasmina Gurung, la jeune locutrice du seke. Durant sept ans elle a documenté la langue à la fois au Népal et à New York, grâce à des dizaines d’heures d’enregistrement, dont beaucoup ont été transcrits et traduits, ainsi qu’un dictionnaire qui continue de s’agrandir. Mais à présent, ses aînés disparaissent peu à peu et emportent leur langue avec eux. Les questions sur l’immigration et l’asile pendent comme une épée de Damoclès au-dessus de la communauté. Les loyers deviennent de plus en plus chers et leur cohésion, pareille à celle d’un village, pourrait ne pas durer.

 

Au cours de la dernière décennie, par hasard, le quartier gurung de Brooklyn est devenu un lieu où des personnes des quatre coins du monde établissent des associations de leur pays natal, des institutions religieuses, des restaurants et une variété d’autres commerces et espaces. Cela donne lieu à la formation de lieux radicalement différents, qui cohabitent les uns à côté des autres. À quelques minutes du « village » seke, on peut entendre des Ghanéens en route vers l’église parler en Twi, des coiffeurs azerbaidjanais parler juhuri et des chauffeurs Uber se réunir autour de kebabs en papotant en ouzbek. Les langues africaines, asiatiques, européennes, caraïbéennes et sud-américaines résonnent dans les carrosseries, les navettes informelles ou « dollar vans », les mosquées, et les bars.

 

Pour tout son potentiel inutilisé, Babel, non pas le mythe biblique mais la réalité contemporaine, s’est réalisé dans les villes telles que New York, à une échelle inimaginable. Il est maintenant temps de le comprendre, l’apprécier et le défendre.

De Ross Perlin

Linguiste et auteur de Language City: The Fight to Preserve Endangered Mother

30 juil. 2026

Cet article a initialement paru dans le magazine National Geographic de juillet 2025."

https://www.nationalgeographic.fr/histoire/linguistique-la-nouvelle-babel-new-york-la-ville-aux-700-langues

#metaglossia 

#metaglossia_mundus 

Language on a prescription label deserves the same regulatory scrutiny given to the drugs themselves

"According to Blank, the language on a prescription label deserves the same regulatory scrutiny given to the drugs themselves.

Drug Topics: Why do you believe that “language on a prescription label deserves the same zero-tolerance standard the industry already demands for the drug itself,” and how does the industry get to that point?

Sharon Blank: The industry would never accept ‘close enough’ on the controlled substance count or an active ingredient’s potency. They both have to be exact and verified. The instructions telling a patient how to take that same drug should be held to the same standard. Let’s take a simple example: if 1 teaspoon 3 times a day gets mistranslated as 3 tablespoons once daily, a single 5-milliliter dose turns into a 45-milliliter dose, 9 times what the prescriber intended, that can very seriously harm a patient.

Getting there takes 4 things. First, the translation has to be checked by a licensed pharmacist who actually speaks the language, not just a general translator or an AI tool. Then second, it should come from a controlled library, so not just typing a string of text but instructions need to be pre-approved. Third, it needs to be built right into the pharmacy software. As I mentioned, this will significantly lessen the burden on the pharmacy flow. And lastly, it has to cover the languages patients really speak, not just a handful that look good in a brochure but that really cover any pharmacy’s language population.

Drug Topics: What support are pharmacists primed to deliver to improve language barriers, and what kind of support is needed on a grander scale that pharmacists aren’t yet positioned to address?

Sharon Blank: Pharmacists are very well positioned to notice the gaps and to counsel. They’re the ones at the counter and watching the patient take the pill bottle from them and just kind of nodding along, knowing full well that there’s very limited comprehension and understanding. [Pharmacists] see that, they notice it, it’s happening, but what they’re not positioned to do, and shouldn’t be expected to do, is to generate a validated translation in real time themselves. It’s not fair to ask of someone who’s first off extremely busy and tight on time but also just not in their area of expertise.

The bigger fix has to come from technology, having a verified translation that gets built right into the pharmacy system, so the correct answer is just what prints on the label. There is no need to think and second guess or question what is being delivered and printed. It also needs industry-wide alignment. We need a solution validated under one state’s framework so it’s not being piecemealed from pharmacy to pharmacy, location by location, chain by chain, but rather having total state mandated adherence."
July 29, 2026
Author(s)Brian Nowosielski, Sharon Blank
https://www.drugtopics.com/view/should-translation-technology-become-a-pharmacy-standard-
#metaglossia
#metaglossia_mundus

Selecting a medical translation company for specialized content
"Language Scientific Addresses What Life Sciences Teams Need From a Medical Translation Company

"La GSMA, en partenariat avec plusieurs gouvernements, universités, opérateurs télécoms et entreprises technologiques, a lancé ATLAS Umoja, une initiative panafricaine destinée à accélérer le développement de modèles d’intelligence artificielle entraînés sur les langues africaines. L’objectif est de combler le retard du continent dans ce domaine en produisant des jeux de données, des modèles linguistiques et des outils capables de mieux comprendre les réalités culturelles et linguistiques africaines. Équipementde communication

 

Aujourd’hui, la majorité des modèles d’IA générative sont principalement entraînés sur des contenus en anglais, en chinois ou dans d’autres langues largement représentées sur Internet. Cette situation limite leur pertinence pour des centaines de millions d’Africains qui utilisent quotidiennement des langues locales. En développant des ressources ouvertes dédiées aux langues africaines, l’alliance espère favoriser l’émergence de services d’IA mieux adaptés à des secteurs tels que l’éducation, la santé, l’agriculture ou encore les services publics.

 

Au-delà de la dimension technologique, l’initiative porte un enjeu de souveraineté numérique. Les données linguistiques constituent aujourd’hui une ressource stratégique pour entraîner les modèles d’intelligence artificielle de demain. En encourageant leur production et leur gouvernance sur le continent, la GSMA et ses partenaires souhaitent permettre à l’Afrique de participer davantage au développement de l’IA plutôt que d’en rester un simple marché d’utilisation."

par AITN 

29 juillet 2026

https://afriqueitnews.com/intelligence-artificielle-ia/gsma-lance-alliance-pour-developper-ia-en-langues-africaines/

#metaglossia 

#metaglossia_mundus 

PhD study develops AI to detect misinformation in South African languages

Dr Seani Rananga, a former NWU PhD student, is developing multilingual AI that can detect misinformation across English, isiZulu and Sepedi

."...The research demonstrates that multilingual AI can identify misleading information across the three languages while highlighting the importance of improving machine translation for low-resource African languages. The study provides a framework that can be adapted to combat misinformation during elections, public emergencies and other events where false information can have serious consequences.

Dr Rananga, a lecturer at the University of Pretoria, said the research was driven by the need to ensure that speakers of South Africa’s indigenous languages are not left behind as AI becomes increasingly important in identifying misinformation.

“South Africa is one of the most linguistically diverse countries in the world, yet most AI systems for detecting misinformation are developed primarily for English. This leaves speakers of indigenous languages at a disadvantage when it comes to identifying false or misleading information online,” she said.

Building trustworthy technologies
Her PhD research used misinformation surrounding the Covid-19 pandemic as a pilot case study to test the multilingual framework.

The pandemic illustrated how quickly false information can spread and the consequences it can have for people’s lives, according to Dr Rananga. While the initial research focused on health misinformation, the methods were designed to be applied across multiple sectors where misinformation presents a public risk.

The study found that multilingual AI can effectively detect misinformation in English, isiZulu and Sepedi while showing that the quality of machine translation plays a significant role in improving AI performance for low-resource languages.

“The framework provides a foundation for building more inclusive and trustworthy AI technologies for South African languages,” Dr Rananga said.

She said the findings could support government departments, fact-checking organisations, social media platforms and communities by enabling earlier detection of misleading information, improving access to credible information and helping people make informed decisions.

Standing proud on the world stage
The study has attracted international recognition. Dr Rananga received the Google PhD Fellowship, one of the world’s leading awards supporting doctoral research in AI. Her work also received the Best Poster Award at the 2025 Deep Learning IndabaX South Africa, earning her funding to present the research at the 2026 Deep Learning Indaba at Pan-Atlantic University in Lagos, Nigeria in August this year.

“Receiving the Google PhD Fellowship was a tremendous honour and a strong validation of the importance of conducting AI research for African languages. It reinforced my commitment to developing technologies that address challenges faced by multilingual and low-resource communities while also creating opportunities to collaborate with leading researchers around the world,” she said.

Dr Seani Rananga is passionate about languages and identifying misinformation. Photo: Supplied
Looking ahead, Dr Rananga plans to expand the research to include more South African languages. Future work will focus on misinformation shared on social media during elections, public emergencies and other high-impact events. She also intends to extend the research to address emerging challenges such as AI-generated deepfakes, hate speech and other forms of harmful online content.

Beyond detection, Dr Rananga aims to develop trustworthy multilingual AI systems, including large language models, retrieval-augmented generation systems and knowledge graphs that can support sectors such as healthcare, education, agriculture, governance and public services.

“My long-term vision is to ensure that African languages are fully represented in the next generation of AI technologies while contributing to safer and more reliable digital spaces for everyone.”"

https://www.citizen.co.za/potchefstroom-herald/news/2026/07/29/phd-study-develops-ai-to-detect-misinformation-in-south-african-languages/
#metaglossia
#metaglossia_mundus

"Avant leur déclin, jusqu’à 75 000 langues auraient coexisté dans le monde

Si nous communiquons aujourd’hui grâce à environ 7 500 langues à travers le monde, une étude suggère qu’il existait bien plus de langues parlées entre l’an 1000 avant notre ère et 1000 de notre ère avant d’amorcer un déclin, notamment avec l’expansion des grands États et empires multinationaux. Cette période, qualifiée d’« âge d’or » linguistique, aurait vu exister entre 20 000 et 75 000 langues différentes.

Le nombre de langues parlées et leur diversité dans le monde diminuent depuis des centaines d’années et continuent de décliner. Les experts estiment qu’au moins quatre langues disparaissent chaque année et que jusqu’à 1 500 langues connues pourraient disparaître d’ici la fin du siècle.

Certaines hypothèses ont avancé que le déclin actuel des langues pourrait être lié à l’essor du colonialisme européen il y a environ 500 ans. Une nouvelle étude suggère cependant que ce déclin s’est déjà produit bien avant. L’étude démontrerait entre autres qu’il faut bien moins que des puissances coloniales pour expliquer le déclin de la diversité linguistique.

« D’une certaine manière, le langage est éphémère — une phrase prononcée disparaît dans les ondes sonores — et pourtant, nos travaux démontrent que nous pouvons utiliser des schémas linguistiques pour reconstituer de grands événements démographiques sur des millénaires », explique Claire Bowern, professeur de linguistique à la Faculté des arts et des sciences de l’université de Yale et co-auteur de l’étude.

Identifier les différents facteurs qui ont façonné notre diversité linguistique actuelle est en effet essentiel pour la compréhension de l’histoire, de la culture et de la cognition humaine. « Nous montrons que le langage n’est pas isolé des courants politiques et que cette dynamique fait partie intégrante de l’histoire humaine depuis bien plus longtemps qu’on ne le pensait. Les courants politiques ne sont pas nécessairement unidirectionnels. L’extinction des langues menacées n’est pas inévitable », ajoute-t-elle dans un communiqué.

Une apogée linguistique entre 1000 av. J.-C. et 1000 apr. J.-C.
Comme il n’existe pas de preuves directes de l’existence de langues avant l’invention de l’écriture il y a environ 6 000 ans, Bowern et ses collègues ont estimé la diversité linguistique au cours de l’Holocène (il y a environ 12 000 ans à aujourd’hui) en combinant des données ethnographiques à des estimations de la taille des populations préhistoriques pour effectuer des modélisations statistiques et socio-informatiques.

Des données ethnographiques provenant de 171 sociétés de chasseurs-cueilleurs et de pêcheurs ont été utilisées pour évaluer la répartition et la taille des populations avant l’essor de l’agriculture et de l’élevage. Ces communautés ont été sélectionnées parce qu’elles sont considérées comme ayant relativement peu changé dans leurs modes de vie et leur mobilité au contact de populations agricoles. Les chercheurs ont également combiné ces estimations avec des analyses indépendantes suggérant que la population humaine mondiale il y a 12 000 ans se situait entre 4,4 et 7 millions d’individus environ.

En supposant que le nombre de langues au début de l’Holocène correspondait au nombre de groupes de langues distincts que la population humaine totale pouvait accueillir à cette époque, les modélisations ont indiqué qu’entre 4 500 et 6 200 langues étaient parlées il y a 12 000 ans.

« Le monde qui a précédé l’agriculture n’était probablement pas exceptionnellement riche en langues », explique Damián Blasi, auteur principal de l’étude et professeur à l’Institut catalan de la recherche et des études avancées (ICREA) et au Centre pour le cerveau et la cognition de l’Université Pompeu Fabra de Barcelone, en Espagne. « Il y avait très probablement moins de langues qu’aujourd’hui. La diversité linguistique s’est ensuite développée parallèlement à la population humaine pendant des milliers d’années. »

Afin d’estimer comment cette diversité a évolué à mesure que les populations augmentaient, les chercheurs ont modélisé des milliers de trajectoires possibles. Mis à part la croissance démographique, l’équipe a également pris en compte l’influence de différents facteurs tels que les guerres, les épidémies et les innovations technologiques.

Les modélisations indiquent systématiquement un « âge d’or » linguistique se situant entre 1000 avant notre ère et 1000 de notre ère, période durant laquelle entre 20 000 et 75 000 langues auraient été parlées dans le monde. « Un résultat surprenant est la relative proximité de ce pic », indique Bowern au Smithsonian Magazine. « La période de plus grande diversité linguistique pourrait ne pas se situer au Paléolithique, mais plutôt à l’époque de l’expansion des États et des premiers empires. »

Carte des langues actuellement parlées et signées dans le monde, colorées selon leur statut de menace. © Blasi et al.

Un âge d’or s’achevant avec l’essor des empires multinationaux
Cet âge d’or se serait achevé avec l’essor des États et des empires multinationaux, les populations dominantes ayant propagé leurs langues et leurs cultures à des populations initialement indépendantes. Le colonialisme européen aurait accentué ce déclin, sans toutefois en être la cause principale, précise l’étude.

D’après Russell Gray, co-auteur de l’étude et directeur du département d’évolution linguistique et culturelle de l’Institut Max Planck d’anthropologie évolutionniste de Leipzig : « les langues que nous connaissons aujourd’hui sont les survivantes d’un goulot d’étranglement historique massif et extrêmement sélectif. »

En effet, une caractéristique linguistique peut être répandue non pas nécessairement parce qu’elle est plus efficace ou plus adaptée à la communication, mais parce qu’elle serait plutôt propagée par des populations en expansion. Les résultats de l’étude sont publiés dans la revue Science.
Source : Science"
Valisoa Rasolofo • 28 juillet 2026
https://trustmyscience.com/avant-leur-declin-jusqua-75-000-langues-auraient-coexiste-dans-le-monde/
#metaglossia
#metaglossia_mundus

Summary
Recent advances in AI inspire visions of universal models of biology. Yet living systems are evolved, emergent processes whose behaviors cannot be inferred from their parts alone. We propose grounding AI in canonical biological processes, constructing data-driven world models with explicit mechanistic links across molecules, cells, and their dynamics in space and time...

Effective datasets for AI on biological processes should reflect the structure of the processes they seek to model. Such datasets must characterize multiple biological scales at an appropriate resolution. Components should ideally be captured within their native context, rather than in isolated experimental settings. For example, genomic regulatory elements are best studied within their epigenomic context and cells within their native tissues. Because canonical processes unfold in space and
Modeling processes by linking multi-scale mechanisms for causal inference
Canonical processes not only structure biological data, but they also invite a synthesis between AI and mechanistic reasoning. Combining AI with causal, perturbation-informed frameworks could allow the construction of process-based “world models”13 of biology. Solving such models will require new computational frameworks14 that integrate modules capable of super-human deduction with those that formalize a “systems biology” approach to biological function and dynamics.15 The goal is not to
Multi-scale computational biology world models: Suggested principles
Biological world models are an open challenge for the coming years. Early insights from current approaches suggest some guidelines for the transformation of building blocks working at one scale into multi-scale frameworks via distillation, standardization, and reconnection. Distillation in the biological domain may be used to reduce redundant and correlative model features and enhance their causal robustness. For example, current genomic transformers that perform well in tasks of predicting
Teaching AI to speak biology
In conclusion, we argue here that scaling in biology will require data and learning tasks anchored in principles and mechanisms that reach beyond current AI success stories. AlphaFold represents only the simplest building block in a hierarchy of models that are needed for approaching the deeper problem of biological computation. Higher-order biological functions should be computed within and around captured canonical processes...
Yonatan Stelzer
Amos Tanay
Cite
https://doi.org/10.1016/j.cell.2026.07.003
https://www.sciencedirect.com/science/article/abs/pii/S0092867426008019
#metaglossia
#metaglossia_mundus