Metaglossia: The Translation World

De l'adoption à l'adaptation : Comment l'Afrique façonne sa propre IA

"Selon la Mise à jour économique pour l’Afrique de la Banque mondiale d'octobre 2026, l'Afrique subsaharienne concentre principalement son portefeuille d'intelligence artificielle dans les phases d’adoption et d’adaptation, tandis que la phase d’avancement (création de modèles frontières) reste marginale.

En raison des coûts élevés et du manque d'infrastructures de calcul, il ressort que la région tire son avantage comparatif de l’IA restreinte/pratique : des outils ciblés, à faible bande passante, conçus pour contourner la connectivité intermittente et la faible alphabétisation numérique.

Pour maximiser l'impact de ces technologies, l'adaptation locale est cruciale. Selon la Banque mondiale, les jeux de données mondiaux négligent largement le continent : environ « 56 % des données d'entraînement en accès libre sont exclusivement en anglais », laissant de côté les 2 000 langues africaines.

Ainsi, « les retombées les plus profondes pour les pays d'Afrique subsaharienne résident dans l'adaptation de l'IA aux données, aux langues et aux problèmes locaux ».

Sur le plan économique, le déploiement de l'IA s'appuie fortement sur la dynamique des startups et le secteur des technologies de l'information, qui a capté « plus de 1,25 milliard de dollars US » en investissements de capital-risque. La technologie se déploie principalement là où les transactions numériques sont déjà ancrées : la fintech, santé, l'agriculture et les services d'identification numérique.

Le rapport souligne que la réussite de ces solutions repose sur des canaux de diffusion accessibles (SMS, USSD, WhatsApp) et sur la présence de « compléments analogiques » indispensables comme l'électricité et les compétences de base.

D’après la Banque mondiale, la réplication à grande échelle exige d'intégrer l'IA aux systèmes existants et de maintenir un modèle de financement pérenne, rappelant que « la pérennité de l'échelle exige de maintenir les performances à mesure que les applications s'étendent à de nouvelles populations »."

Fodé Bakary Camara
15 Octobre 2026
https://www.pressafrik.com/De-l-adoption-a-l-adaptation-Comment-l-Afrique-faconne-sa-propre-IA_a313107.amp.html
#metaglossia
#metaglossia_mundus

Every person on the planet speaks with a vocal tract that is slightly different from everyone else’s
Differences in human vocal anatomy may help explain why languages around the world sound so different.
Human language diversity may have surprising biological roots
Scientists discover how differences in human vocal tract anatomy may help explain why the world's languages sound so different.

Every person on the planet speaks with a vocal tract that is slightly different from everyone else’s.

The tube of flesh, bone, and cartilage that turns breath into speech comes in endless shapes and sizes. No two people use it in quite the same way.

This individual variation was long dismissed as irrelevant noise by linguists.

A new study suggests it may actually be one of the hidden influences behind the astonishing diversity of the world’s spoken languages.

The research was led by Steven Moran and Tiena Danner from the University of Neuchâtel. It offers a biologically grounded explanation for a puzzle that has fascinated scholars for centuries.

Why do the roughly 7,000 languages currently spoken on Earth sound so different from one another?

Scanning throats in real time
To get at this question, the researchers took an unusual approach.

They analyzed data from 32 American English speakers – 16 men and 16 women – using real-time MRI scans taken while each person was actively speaking.

The scans, drawn from the USC 75-Speaker Speech MRI Database, captured the shape of each speaker’s vocal tract at the precise moment they were producing specific vowel sounds. What the imaging revealed was striking.

Even when producing the same vowel, speakers moved their throats, tongues, and lips in remarkably different ways.

Some had short, rounded palates; others had longer, flatter ones. Each person’s tongue adapted to their specific anatomy.

For instance, a rounded palate produced a more rounded tongue shape. These differences left measurable acoustic fingerprints in the sounds they produced.

Speakers compensate for their anatomical differences, the team found. They adjust their articulatory movements to converge on the same target sound.

But the compensation is imperfect. Vocal tract anatomy still influences the acoustic signal, leaving each speaker sounding slightly distinct even when pronouncing identical sounds.

The personal signature in your voice
This is not entirely surprising to anyone who has ever recognized a friend’s voice on the phone. What the study does is quantify the mechanism behind it, and then take the logic one step further.

The researchers compared the variation found among English speakers to phonetic variation recorded across nine other languages.

These languages were Basque, Bulgarian, Hindi, Italian, Portuguese, Romanian, Tamil, Turkish, Russian, and Vietnamese.

They found that the pattern of variation within any single language mirrors the pattern of variation between languages.

Thus, the acoustic spread found among English speakers looks strikingly similar to that found comparing English to Hindi to Turkish.

The implication is significant. The seeds of phonetic differences between languages are already present in the variation between individual speakers within a single community.

A biological baseline for language change
The researchers draw on a principle from evolutionary biology: variation within a population provides the raw material for differences to emerge between populations.

The same appears to be true of language.

Phonetic variation among individual speakers, influenced by their vocal tract anatomy, may contribute to the differences that emerge between languages.

Under this model, the physical characteristics of speakers may help shape how languages develop over time.

This matters because linguists have long treated individual phonetic variation as noise obscuring the underlying structure of a language.

Speech recognition systems discard it and phylogenetic analyses of language families ignore it. The new study argues that doing so leaves out important information.

The authors suggest their model can serve as a null hypothesis for studying phonetic evolution. If languages drift apart naturally because speakers vary anatomically, that gives researchers a baseline.

Any deviation from that baseline, such as languages converging or diverging faster than expected, could signal that other forces are at work, including history, culture, or social pressure.

Biology gives you the baseline; anything on top of it is human history.

From individual to language
The path from an individual’s particular palate shape to two communities eventually being unable to understand each other is long and winding.

The researchers do not claim more than their data supports. But the direction the evidence points is intriguing.

Shared ancestry, diet, or developmental environment can all nudge a community’s vocal anatomy in a common direction.

And if enough speakers share the same anatomical tendencies, those tendencies may quietly shape how the whole community sounds.

Over generations, as communities diverge and cultural forces push sounds in different directions, small anatomical predispositions could compound into substantial linguistic differences.

The study is published in the journal PLOS Biology."
ByAndrei Ionescu
Staff Writer
October 14, 2026
https://www.earth.com/science/human-language-diversity-may-have-surprising-biological-roots/
#metaglossia
#metaglossia_mundus

"
ASU lab addresses need for culturally sensitive multilingual AI learning tools
AI and Multilingualism Lab allows for development of ethical, transparent and socially responsible language learning tools

Faculty and students collaborate at the AI and Multilingualism Lab, housed within ASU's School of International Letters and Cultures. Photo courtesy of Marta Tecedor Cabrero

By Thit Thit T Win |
October 14, 2026
When a Spanish-Quechua speaker in Ecuador switches between the two languages while searching for health information online, rather than recognizing their natural bilingual communication, AI flags it as broken language.

That's because the AI behind most language tools was not built with them in mind. It was trained on standard, monolingual data.

"If an AI system does not understand linguistic variation, it can mistake difference for deficiency," said Olga Kellert, associate professor at Arizona State University's School of International Letters and Cultures, a unit within The College of Liberal Arts and Sciences.

Kellert leads a project that applies AI and computational methods to health communication for Spanish-Quechua speakers in Ecuador and Peru. Originally supported by the German Research Foundation, the research examines how bilingual populations access and interpret health information, including communication about COVID-19.

"Most digital language tools are developed for a relatively small number of globally dominant languages," Kellert said. "Many Indigenous languages, including Quechua, have far fewer computational resources."

Research at ASU's AI and Multilingualism Lab, where Kellert's project is housed, works to close that gap.

Starting with how people actually speak
Rather than training AI on textbook language, Kellert's team works with real interactions collected from classrooms and communities.

"Language learning happens in real interactions, and those interactions are messy, variable and deeply human," she said. "Students do not always produce the kind of language that appears in textbooks. They hesitate, repair what they say, switch languages, use regional varieties and develop in different ways."

This is deeply interdisciplinary work. Kellert collaborates with ASU's School of Computing and Augmented Intelligence, supervising a master's student and a PhD student. The team also works with the University of A Coruña in Spain.

"As linguists, we ask: What exactly is the speaker doing here? Is this an error, a dialectal form, code-switching or a normal feature of bilingual communication?" Kellert said. "Computer scientists ask how we can represent that information computationally and build systems that can detect it reliably."

Students in the lab work with real linguistic data at every stage, from collection and annotation to analysis with Python and NLP tools.

"We want students to see that they don't have to choose between the humanities and technology," she said.

What a language lab has to do with AI
The AI and Multilingualism Lab is directed by Marta Tecedor Cabrero, associate professor of Spanish, and it brings together faculty across the School of International Letters and Cultures who have worked for years in computer-assisted language learning, computational linguistics and digital humanities.

A graduate student pilot tests an AI language tool with the assistance of an undergraduate student. Photo courtesy of Marta Tecedor Cabrero
"The lab is a space for thinking critically about the use of AI as it relates to multilingual practices," Tecedor Cabrero said. "Language permeates every area of human knowledge, and we in the humanities play an important role in contributing to our collective understanding of this new technology — because we are language and culture experts, not the machines."

The lab's projects range from computational analysis of ancient poetry to research on how literature has shaped human responses to artificial beings, to an international collaboration training AI to analyze a century of Latin American immigration laws.

Its most student-facing tool, Language Buddy, is an AI conversation partner integrated into Canvas. Developed through the ASU AI Innovation Challenge and now being expanded with EdPlus, it has been tested in 12 Spanish 101 sections, with 334 of 679 enrolled students interacting with it across 801 conversations.

Tecedor's own research has produced a counterintuitive finding: Students who only practiced with human partners ended up more willing to try speaking with an AI bot than students who had actually been using one all along.

"We think it may reflect what we call the 'AI hype versus reality effect,'" Tecedor said. "Students in the AI condition had actual firsthand experiences on which to base their responses. Their smaller increase suggests a more realistic understanding of what using an AI avatar bot to practice a second language really entails."

Teachers are not going anywhere
Tecedor said the biggest misconception is that AI will replace human instructors. She sees a parallel that reaches back centuries.

"In medieval times, knowledge was stored in books that were written out and copied manually, so access to knowledge was severely limited," she said. "The invention of the printing press changed learning drastically. We have testimonials from those times of scholars who saw the printing press as a threat to learning for reasons very similar to those we hear today about AI. But just as teachers did not disappear then, I don't think they will disappear today. Because learning is an inherently social activity."

The lab is seeking undergraduate researchers through Barrett College Fellows and the OURS program.

"Facilitating student-faculty collaborations is one of the main goals of the lab," Tecedor said. "Some students are collaborating with us directly on our research, which includes presenting at national and international conferences and co-authoring papers."

Learn more about the AI and Multilingualism Lab at silc.asu.edu/ai-multilingualism-lab."
https://news.asu.edu/20261014-arts-humanities-and-education-asus-ai-and-multilingualism-lab-finds-ai-falls-short
#metaglossia
#metaglossia_mundus

"Why Language Models Hallucinate
Events section menu
Events
Colloquiums & Seminars
Public Lectures
User Meetings
October 15, 2026
2:30 PM – 3:30 PM
Building 240
Conference Room 4301
Also Online
Speaker: Santosh Vempala, Georgia Institute of Technology
LANS Seminar
Abstract: Large language models often guess when uncertain, producing plausible yet incorrect statements instead of admitting uncertainty. Such ​“hallucinations” persist even in state-of-the-art systems. We analyze this phenomenon from a mathematical perspective and find that the statistical pressures of the next-word prediction training pipeline induce hallucinations and current evaluation procedures reward guessing over acknowledging uncertainty. We also propose ​“open rubric” evaluations with explicit error penalties, providing a practical path to reliable LLMS by aligning their incentives. The talk will be fact-based, and the speaker will readily admit ignorance.

Joint work with Adam T. Kalai, Ofer Nachum and Eddie Zhang (Nature, 2026).

Bio: Santosh Vempala is Frederick Storey II Chair of Computing and Distinguished Professor in the School of Computer Science, with courtesy appointments in the schools of Mathematics and Industrial and Systems Engineering (ISyE). He served as the founding director of the Algorithms and Randomness Center and ThinkTank (2006-2011), and is currently the director of GT’s oldest interdisciplinary PhD program --- Algorithms, Combinatorics and Optimization. His research interests include algorithms, randomness, high-dimensional geometry, brain and computing-for-good (C4G). He graduated from CMU in 1997 advised by Avrim Blum and was on the MIT faculty until 2007 except for a year as a Miller Fellow at UC Berkeley. Vempala is also a Sloan, Guggenheim, ACM, AMS and generally excitable Fellow, especially when a phenomenon that appears complex from one perspective turns out to be simple from another. In recent years, he has been trying to understand the limits of sampling and optimization algorithms and building a computational theory of brain.

Series: See all upcoming talks at the LANS Seminars page.

Tags: High Performance Computing, Artificial Intelligence, Machine Learning
Join via Zoom"

https://www.anl.gov/event/why-language-models-hallucinate
#metaglossia
#metaglossia_mundus

Google's translation service handles one trillion words per month
That doesn't count the others: DeepL, Microsoft, Claude, ChatGPT and Gemini, not to mention made-to-measure corporate translation systems.

"How long will humans be needed for translation?...

In April, Google said people are now translating one trillion words a month with Google Translate.

That doesn't count the others: DeepL, Microsoft, Claude, ChatGPT and Gemini, not to mention made-to-measure corporate translation systems.

Direct sales of machine translation (i.e. outside general AI subscriptions) were worth an estimated $1.4bn (£1bn) last year.

AI translation is the "right call" in many cases says John Zimmerman
"For plenty of use cases, raw AI translation is now the right call: multiplying your costs by 100 to catch the last few percent of errors isn't rational if you're translating internal documents or product listings," says John Zimmerman, founder of OKA (Open Knowledge Association), a Swiss non-profit that translates Wikipedia articles into other languages.

"For an encyclopedia it's different. Wikipedia runs on trust. A wrong fact can sit there for years and get cited by journalists and AI systems," he says.

With more than 10,000 published pieces, mostly into Spanish, Portuguese and English, OKA is, Zimmerman says, the single largest contributor of translated articles to the collaborative online encyclopedia.

Which is something it absolutely could not have become without AI.

"Translating 10,000 articles the traditional way would cost millions. In our process, the AI produces the draft, then a human translator checks every sentence," he explains.

But that human input is required less and less.

"The machine output is now good enough that skipping the human entirely has become a serious option, and a real temptation across the industry. Two or three years ago, with tools like DeepL, our translators still rewrote a good chunk of every draft.

"Today the prose often doesn't need fixing at all. So, the question is no longer whether AI translation is good, it clearly is, but whether it's good enough to remove the human," says Zimmerman.

Not yet, translators argue...

Last year OKA ran a study, funded by Wikimedia. It studied 119 articles, 10 language pairs and more than 1,000 hours of translator work.

"27% of the AI-generated text still got corrected before publication. Three quarters of those corrections were errors the AI introduced, and 5 to 6% of errors changed the meaning of the original text," says Zimmerman.

Eradicating those errors is likely to be challenging...

This is because probability is a crucial part of the way AI translates each word rather than another and what is probably correct is not the same as what is correct. Especially in the realm of something as subtle and surprising as language...

But perhaps the big moment will be when AI can show doubt..."

John Laurenson
Business reporter, Paris
https://www.bbc.com/news/articles/c1wxdjx27ypo
#metaglossia
#metaglossia_mundus

Face à l'IA, 37 % des traducteurs littéraires interrogés envisageraient de changer de métier
Une enquête européenne menée auprès de 1785 traducteurs littéraires révèle que 37 % ont songé à changer de profession, tandis que près de 30 % déclarent subir les conséquences économiques de l'intelligence artificielle. Moins de commandes, rémunérations dégradées, multiplication des propositions de correction de textes produits par des machines : à la Foire de Francfort, ces résultats ont relancé les discussions sur les garanties que les éditeurs et les organismes de soutien peuvent apporter aux professionnels.

"Le Conseil européen des associations de traducteurs littéraires (CEATL) a rendu publics, le 5 octobre 2026, les principaux résultats de son enquête sur l'intelligence artificielle générative. Le constat est préoccupant : plus d'un tiers des répondants ont envisagé d'abandonner leur métier, principalement en raison de l'essor de cette technologie et de la dégradation de leurs perspectives professionnelles.

Une précision importe : il ne s'agit pas de départs effectivement constatés. Selon le compte rendu de l'association tchèque Obec překladatelů, qui avait présenté des résultats préliminaires lors de l'assemblée du CEATL à Oslo, en mai dernier, les 37 % concernent des professionnels ayant envisagé cette reconversion au cours des trois dernières années.

L'étude, conduite auprès de traducteurs littéraires entre le 12 et le 31 janvier 2026, a été élaborée par le groupe de travail du CEATL consacré à l'IA, avec la collaboration de chercheurs de l'Université ouverte de Catalogne et de l'Université de Zagreb. Elle prolonge deux précédentes investigations réalisées en 2023 et 2024, à une époque où les conséquences professionnelles des outils génératifs commençaient seulement à être documentées.

La postédition, un travail moins payé et moins reconnu
À la question des départs envisagés s'ajoutent des difficultés économiques déjà ressenties. Environ 30 % des participants font état d'une réduction du nombre de commandes ou de clients, de perspectives perdues et d'un déplacement vers des prestations moins rémunératrices, qu'ils attribuent à l'intelligence artificielle.

Ces déclarations ne permettent pas, à elles seules, de mesurer la contraction du marché européen. Elles éclairent néanmoins les transformations que les intéressés disent observer dans leurs relations avec les donneurs d'ordre.

L'une d'elles concerne la postédition, autrement dit la révision d'une traduction produite automatiquement. 32,5 % des répondants ont reçu des propositions de ce type. Parmi eux, 67,3 % déclarent les avoir refusées.

Le CEATL avance plusieurs raisons : les délais accordés seraient généralement plus courts, les tarifs moins avantageux et la contribution des professionnels insuffisamment reconnue, alors même que les exigences de qualité resteraient élevées. À la rémunération s'ajoute donc la question de la nature du travail demandé : non plus élaborer un texte dans une autre langue, mais intervenir sur une production dont les choix initiaux échappent au traducteur.

Le malaise dépasse ces seules propositions commerciales. Plus de 77 % des personnes interrogées expriment une opinion négative sur le développement de l'IA générative dans la traduction et l'édition.

Pour autant, ces résultats ne signifient pas que l'ensemble des professionnels rejettent indistinctement tous les usages informatiques. Le communiqué du CEATL insiste sur les conséquences des dispositifs génératifs pour l'exercice de la profession, sans fournir encore l'intégralité des données permettant d'examiner les pratiques selon les différents outils employés.

La prudence méthodologique reste nécessaire : les 1 785 réponses proviennent d'une participation volontaire, et non d'un échantillon représentatif de tous les traducteurs européens. Près de la moitié des participants indiquent d'ailleurs que la traduction littéraire ne constitue pas leur principale source de revenus. Le CEATL fédère, pour sa part, 38 associations réparties dans 30 pays et représentant environ 12 000 professionnels.

Le droit d'auteur, autre ligne de fracture
L'utilisation des traductions existantes pour entraîner les systèmes génératifs nourrit également de fortes oppositions. 55 % des répondants refusent que leurs textes soient employés à cette fin, quelles que soient les conditions proposées.

Environ un tiers accepterait cependant d'envisager une telle exploitation, à condition qu'elle repose sur un consentement explicite, une rémunération équitable et des modalités précisément définies. Le CEATL mentionne notamment la possibilité de limiter certains usages à des activités non commerciales.

À LIRE - Intelligence artificielle et traduction, l'impossible dialogue ?

La distinction est essentielle : la question ne porte plus seulement sur l'introduction d'un logiciel dans le processus éditorial, mais sur la réutilisation d'œuvres protégées et les droits de ceux qui les ont créées.

En France, l'Association des traducteurs littéraires de France (ATLF) a déjà adapté son contrat-type. Dans sa version actualisée en décembre 2024, elle recommande trois dispositions : exclure le recours à l'IA générative pour établir la traduction, interdire son utilisation dans l'exploitation des droits cédés et formaliser l'opposition à l'emploi des textes pour entraîner des modèles.

Cette dernière clause impose à l'éditeur une obligation de moyens pour faire valoir le refus du traducteur. Le dispositif porte également sur les exploitations dérivées : l'association donne ainsi l'exemple d'une adaptation en livre audio qui ne devrait pas recourir à une voix générée artificiellement.

Les pratiques contractuelles demeurent toutefois inégales. Dans son analyse d'une enquête menée auprès de ses adhérents en mars 2025, l'ATLF observait que 73 % des contrats comportant une clause relative à l'IA interdisaient l'utilisation de logiciels de traduction automatique fondés sur cette technologie. Seuls 26 % prévoyaient une opposition à la fouille de textes et de données.

L'étude française relevait parallèlement une baisse du niveau de vie, la multiplication de tâches annexes peu rémunérées et des difficultés persistantes dans la négociation avec les éditeurs. Les prestations complémentaires, telles que les fiches de lecture, les rencontres publiques ou les ateliers, ne représentaient que 4 % des revenus des répondants. Des fragilités qui dépassent donc la seule question technologique.

En Allemagne, les lecteurs se montrent bien plus favorables
Ces inquiétudes professionnelles contrastent avec la perception du public allemand. Selon un sondage publié le 6 octobre par l'association numérique Bitkom, 79 % des personnes interrogées approuvent l'utilisation de l'IA pour traduire des livres.

L'adhésion diminue sensiblement dès qu'il est question de création littéraire. Seuls 34 % jugent acceptable qu'un écrivain recoure à l'intelligence artificielle pour l'aider à rédiger son ouvrage, et 29 % accepteraient de lire un livre entièrement écrit par une machine.

Les réserves apparaissent aussi sur la transparence : 72 % souhaitent que l'utilisation de l'IA dans les livres soit signalée. Par ailleurs, 65 % redoutent que cette technologie puisse remplacer les auteurs humains.

Réalisée par téléphone auprès de 1 006 personnes âgées de 16 ans et plus, l'enquête allemande est présentée comme représentative de cette population. Elle ne peut donc être directement comparée au questionnaire du CEATL, réservé aux professionnels volontaires.

Elle révèle néanmoins une différence de perception : le recours à des systèmes automatisés pour produire une traduction rencontre, dans le public allemand interrogé, une acceptation nettement supérieure à celle exprimée pour la création intégrale d'un texte littéraire.

Les aides européennes face aux pratiques des éditeurs
La présentation des travaux du CEATL à Francfort, le 7 octobre, intervenait dans un contexte où plusieurs organismes de soutien à la traduction cherchent précisément à renforcer les garanties accordées aux professionnels. Lors d'une rencontre consacrée au réseau européen ENLIT, les participants ont détaillé le fonctionnement des aides publiques destinées à favoriser la circulation internationale des œuvres.

Le réseau réunit 30 membres issus de 25 pays. Il ne distribue pas directement de subventions, mais permet aux institutions qui en accordent de confronter leurs pratiques. En 2025, leurs dispositifs ont soutenu plus de 4 000 livres traduits dans 60 langues, pour une aide moyenne d'environ 2 900 euros par titre.

D'après les chiffres présentés à Francfort, le Centre national du livre (CNL) français affichait le montant total le plus élevé, avec environ 1,75 million d'euros, devant la Tchéquie, à quelque 1,58 million d'euros.

L'effort tchèque s'inscrit notamment dans la préparation de sa présence comme invitée d'honneur de la Foire du livre de Francfort 2026. Martin Krafl, responsable du Centre littéraire tchèque, a indiqué que son pays avait soutenu, cette année, la traduction de 194 ouvrages vers l'allemand, mais également des publications dans d'autres langues.

L'attribution d'une subvention ne suffit cependant pas toujours à garantir que le professionnel soit rémunéré. Martin Krafl a évoqué des éditeurs qui, une fois l'aide obtenue, ne règlent pas leurs traducteurs, en invoquant des problèmes de trésorerie. Les organismes financeurs sont alors contraints de renforcer leurs contrôles.

En Flandre et en Catalogne, les institutions concernées demandent désormais aux éditeurs de fournir une preuve du paiement du traducteur avant le versement des aides.

La France dispose également de mécanismes de vérification. Pour ses subventions à la traduction d'ouvrages français vers les langues étrangères, le CNL exige notamment, après publication, un justificatif du règlement effectué. En l'absence des pièces demandées dans les délais prévus, la subvention n'est pas versée. Le montant de l'aide peut aussi être réajusté lorsque les frais de traduction réellement engagés sont inférieurs d'au moins 10 % à ceux annoncés.

Des clauses pour garantir une traduction humaine
Reste la question de l'intelligence artificielle elle-même. Elise Vanoosthuyse, représentante de Flanders Literature, a expliqué que son organisme réfléchissait à l'introduction d'une clause spécifique dans les conventions de subvention.

Tout en reconnaissant que l'IA pouvait servir d'outil, à la manière d'un dictionnaire, elle a insisté sur ses limites en matière de création : « Pour l'instant, l'IA ne sait pas faire certaines choses, comme restituer l'humour, l'émotion ou trouver la voix juste. »

Du côté de l'Institut Ramon Llull, qui accompagne notamment la traduction d'œuvres catalanes, Izaskun Arretxe a indiqué qu'aucune disposition de ce type n'avait encore été introduite, mais que le futur dispositif devrait préciser que les traductions soutenues sont réalisées par des êtres humains.

L'institution catalane prépare parallèlement, pour la fin novembre, un premier forum consacré aux traducteurs littéraires du catalan. Quelque 150 professionnels travaillant vers d'autres langues doivent se retrouver à Barcelone pendant trois jours, afin de développer leurs échanges et de rencontrer des éditeurs catalans."

Clément Solym
Contact : cs@actualitte.com
Le 12/10/2026

https://actualitte.com/article/134547/auteurs/face-a-l-ia-37-des-traducteurs-litteraires-interroges-envisageraient-de-changer-de-metier
#metaglossia
#metaglossia_mundus

Best Indigenous Languages Graduate Paper Award / Prix du meilleur article de fin d’études sur les langues autochtones

"Call for submissions 2027
Canadian Modern Language Review: Best Indigenous Languages Graduate Paper
Deadline: April 30, 2027

The Best Indigenous Languages Graduate Paper Award is an annual competition open to all Indigenous graduate students within two years of graduation with a master’s or doctoral degree.

Submissions are to incorporate an explicit focus on Indigenous languages in contexts of teaching, learning, and other forms of community engagement. While we encourage papers with a pedagogical focus, including curriculum, teaching methods, and learning processes, we also welcome papers that consider the broader context of colonial and decolonial trajectories in language use and reclamation and the role of community, cultural, and policy initiatives in Indigenous language resurgence. Papers with Indigenous authorship or co-authorship are encouraged and can be published as "open access."

Submitted papers must adhere to all the requirements of regular submissions and describe original research, either qualitative or quantitative. Articles should be between 7,000 and 10,000 words, inclusive of references, figures, tables, notes and appendices and can be written in English, French, First Nations, Metis and Inuit languages, or a combination of these.

Submissions are due on April 30 of each year and will be evaluated by the editors and members of the journal’s Indigenous Languages Advisory Committee. The winning submission is normally published in the following year.

For more information please consult the CMLR Guidelines for Authors.

Meike Wernicke
Michael Zuniga
Editors-in-chief, The Canadian Modern Language Review"
https://utppublishing.com/journals/cmlr/best-indigenous-languages-graduate-paper-award
#metaglossia
#metaglossia_mundus

Putting the reader first
Estonia's first plain language standard took effect in early October, offering guidance on writing clear, reader-focused public communications...
Under the international definition, a text is written in plain language when its intended readers can find the information they need, understand it and act on it.

"Putting the reader first is the most important thing," Põldma said. "We need to think about who we're writing for or who we're making a program for. That determines the words we use, the sentences we write and what the text should look like

"Estonia's new plain language standard aims to make public writing clearer

Estonia's first plain language standard took effect in early October, offering guidance on writing clear, reader-focused public communications.

Adopted ahead of International Plain Language Day on Tuesday, the standard aims to make public-sector writing clearer and communication more effective.

Kristel Ress, who chairs the technical committee that developed the standard, said writers should first consider who will read a text and in what circumstances.

Instructions for a fisherman reading on a phone in a cold boat, for example, need to account for very different conditions than something read at home or standing at a bus stop.

Ress nonetheless emphasized that the standard is just one tool, offering broad principles rather than detailed rules on word choice or grammar.

"The advantage of the standard is that, for the first time, all this advice has been brought together in one concise guide," she explained.

This guide, she continued, can help anyone who writes for work, including editors, teachers and researchers — but also office workers and even students.

"This isn't just for language professionals anymore," Ress said.

Putting the reader first

ERR language editor and adviser Ester Põldma said the guidance will be particularly useful for people writing online, where brevity and clarity are essential.

"Readers should be able to understand on the first read what a sentence is trying to convey," Põldma said.

Under the international definition, a text is written in plain language when its intended readers can find the information they need, understand it and act on it.

"Putting the reader first is the most important thing," Põldma said. "We need to think about who we're writing for or who we're making a program for. That determines the words we use, the sentences we write and what the text should look like.":l
Indrek Ojamets
13.10.2026
https://news.err.ee/1610164957/estonia-s-new-plain-language-standard-aims-to-make-public-writing-clearer
#metaglossia
#metaglossia_mundus

"Use Crossword Generator to create challenging custom puzzles in 27 languages
Try to solve an exclusive 'Popular Science' crossword, then make your own.

Crossword Generator lets users create 13x13 grid puzzles for free.

Crossword puzzles range from those easy elementary school-level varieties up to nightmarish labyrinths of text. But crafting even the most simple versions remains its own artform. Seamlessly combining dozens of answers into a coherent, functional grid often requires years of practice, but digital strategist and creator Tex Jernigan has designed an impressive tool to level the playing field, so to speak: Crossword Generator.

“Making my own puzzles is something I wanted to do for my neighborhood coffee shop, and [I] started working on it many years ago,” Jernigan, who also recently debuted a vintage Blockbuster VHS sleeve maker, tells Popular Science. “I thought I would be able to have a huge list of theme words, but it can be quite difficult to get more than a small handful of theme words to work.”

Crossword Generator’s creator designed a custom edition for ‘Popular Science’ available for free online. Credit: Tex Jurnigan
To use Crossword Generator, creators first pick a grid size, then choose either a pattern with pregenerated or manually entered black squares. Next, creators have the option to fill the grid with a few specific words, or simply let the program create an entire puzzle itself. Either way, Crossword Generator employs a constraint solver to scour a database of hundreds of thousands of words, then starts plugging in potential options on a trial-and-error basis. Preference is given to more common, everyday words instead of particularly obtuse selections. Any words entered by a user remain locked in during the generation process, and the prom works around those as it fills in the blanks. Crossword Generator starts in English, but supports 27 other languages including Spanish, Russian, and Japanese.

While Jernigan’s constraint solver doesn’t rely on artificial intelligence, another generative program does use Google Gemini to crossmatch every word with a dictionary-based clue. That said, anyone can simply erase those and create their own hints to ensure their customized crosswords are as human-centric as possible.

“In my experience, AI isn’t very good at building a crossword from scratch, because it can’t reliably make every word cross,” Jernigan explained on his website. “It works much better as a partner to the solver.”

Crossword Generator is free for creating and printing up to 13×13 puzzles, but a subscription service opens up New York Times-style grids ranging from the 15×15 through 23×23 Sunday editions. But if solving crosswords is more your thing than creating them, you can start working on Jernigan’s exclusive Popular Science-themed puzzle."
Andrew Paul
Oct 13, 2026 3:09 PM EDT
https://www.popsci.com/technology/crossword-puzzle-generator/
#metaglossia
#metaglossia_mundus

An award for translators from Ukrainian into other languages

"Open Call for the 2026 Drahomán Prize
october 12, 2026

On October 12, the founding organizations of the Drahomán Prize, an award for translators from Ukrainian into other languages, announced an open call for 2026.

Based on the conditions of the Prize, all translators who have at least one translated and published literary or documentary work from Ukrainian language, can be nominated for the Prize. Literary and documentary works include prose, poetry, drama, and non-fiction (essays, reportage, interviews, biographies, memoirs). At the same time, anthologies of translated texts by three or more translators are not eligible for submission.

Translators can be nominated for the Prize by foreign cultural and academic institutions, diplomatic missions, publishing houses, creative associations, and members of the Chapter. The nominated book of the translator must be published by a foreign publisher within two years prior to the award year and should be compatible with values and principles of the founding organizations. Nomination of a tandem of translators who worked together on translating the same text is possible. Laureates of the Prize cannot be nominated for the second time.

The submission includes a filled-out application form as well as electronic and hard copies of the nominated book in Ukrainian and in the language of translation. Applications for 2026 will be accepted from October 12 to December 4.

Please send electronic copies to drahoman.prize@gmail.com. Details for sending paper copies of publications:

Recipient: Angelina Kucherenko
Telephone number: 0 (63) 704 75 00
Email: info@pen.org.ua
Postcode: 04052
Address: 14a Lukyanivska Street, Office 1, Kyiv, Ukraine
The long list of nominees for the 2026 Drahomán Prize will be announced no later than December 10. The winner of the Prize will be announced at a gala ceremony in spring 2027. The winner will receive a statuette created by a Ukrainian artist, a cash prize of 3,000 euros (after tax), as well as further opportunities to work on and promote their art.

The German translator Claudia Dathe became the first laureate of the Prize in 2020. Also, Imadeddine Raef, a translator from Ukrainian into Arabic, was awarded the Chapter’s Special Honor "for the excellence in translation and promotion of Ukrainian classical literature."

The Polish translator, writer, and literary critic Bohdan Zadura became the laureate of the 2021 Drahomán Prize. The Special Honor of the Chapter "for the filigree translation of the classics of Ukrainian literary modernism and contribution to the promotion of Ukrainian literature in Europe" was awarded to Tobias Wals, a Ukrainian-to-Dutch translator.

The French translator Iryna Dmytrychyn became the laureate of the 2022 Drahomán Prize. The Special Honor of the Chapter "for the particular contribution in translating and promoting Ukrainian literature, and in developing the Ukrainian studies in Italy" was awarded to Giovanna Brogi, a translator from Ukrainian into Italian.

The Polish translator Katarzyna Kotyńska was named the 2023 laureate of the Drahomán Prize. The Special Honor of the Chapter "for building cultural bridges between Ukraine and Georgia" was awarded to Raul Chilachava, a translator from Ukrainian into Georgian.

The Finnish translator Eero Balk has been awarded the 2024 Drahomán Prize. The Special Jury Commendation for the high professionalism and humanism in developing the Ukrainian-Lithuanian literary dialogue was given to a translator from Ukrainian into Lithuanian, Vytautas Dekšnys.

The Ukrainian-American translator and poet Nina Murray has been awarded the 2025 Drahomán Prize. The Special Jury Commendation "for her profound contribution to building a shared context between two distant cultures" was given to Mridula Ghosh, a translator from Ukrainian into Bengali.

***

Ukrainian Institute is a state institution dedicated to spotlighting Ukrainian culture globally and nurturing profound cultural ties with other countries and peoples. Since 2018, the Institute has worked worldwide with the head office in Kyiv. In 2023, its first branches opened in Europe: Berlin, Germany, and Paris, France. The Ukrainian Institute is affiliated with Ukraine’s Ministry of Foreign Affairs.

PEN Ukraine is a cultural and human rights NGO uniting Ukrainian intellectuals – writers, journalists, scholars, publishers, translators, human rights defenders, culture managers. With over 170 members, it is one of 146 national centers of PEN International. It is a co-founder of the Vasyl Stus Prize, the Yuri Shevelyov Prize, and the George Gongadze Prize.

Ukrainian Book Institute is a government entity affiliated to the Ministry of Culture and Information Policy of Ukraine. Its mission is to develop state policy in the book sector, promote book reading in Ukraine, support book publishing, encourage translation activity, and popularize Ukrainian literature abroad."
https://pen.org.ua/en/rozpochavsya-konkurs-na-zdobuttya-premiyi-drahom-n-prize-za-2026-rik
#metaglossia
#metaglossia_mundus

"Lost In Translation: How Language Changes Which Brand AI Recommends
Your English-only AI visibility checks can miss how brand recommendations change when AI answers the same buying questions in five different languages.

Reza Moaiandin

Chances are you’ve already got some idea of your brand’s visibility in AI responses. But I’m willing to bet those visibility checks were carried out in just one language: English.

If a customer in Singapore asks an AI chatbot a buying question in Simplified Chinese, and a customer in Kazakhstan asks the same question of the same AI in Russian, would they get the same recommendation? What about a customer in Riyadh asking in Arabic or a customer in Los Angeles asking in Persian?

These aren’t hypothetical customers. According to Statcounter’s August 2026 figures, ChatGPT was responsible for roughly three-quarters of website visits that came from AI chatbots in Saudi Arabia, Kazakhstan, and Singapore. Add in Claude and Gemini, and these three U.S.-based models accounted for more than 90% of AI referrals in each market.

Of course, you don’t need to leave the U.S. to find customers whose first language isn’t English. More than half of Los Angeles County residents speak another language at home, including around 74,000 Persian speakers, according to the latest Census estimates. I’d be surprised if they all switched to English every time they opened ChatGPT.

But when somebody asks an AI chatbot a question in another language, the response won’t simply be a translation of the answer it would give in English. The AI reasons and composes its response in that language. And that brings a whole slew of cultural values and other factors to bear on how the model responds.

Why The Choice Of Language Matters In AI
I’m not the first person to write for Search Engine Journal on how brand visibility and messaging in AI can come across differently in languages beyond English. In April, Duane Forrester explained why AI visibility strategies built in English don’t transfer easily to other markets. For a start, ChatGPT and Gemini aren’t accessible in China at all, while Claude isn’t officially offered in mainland China, Russia, or Iran. Instead, your brand’s AI visibility in some countries may depend on regional, homegrown models, such as Baidu’s ERNIE in China, Naver in South Korea, Mistral in France, and so on. As Forrester explains, each of these models evolved from a different starting point entirely: a national corpus, a specific cultural identity, and in some cases a government mandate. If you haven’t read his article, I’d bookmark it now to read after this one.

However, as we’ve seen, Western AI models like Claude, ChatGPT, and Gemini are still widely used in plenty of locations where other languages are spoken, not to mention all the diaspora communities around the world.

In July, Anthropic published a report on Claude’s values across languages and models. By values, we’re talking about the factors that shape responses where there is no definitive right answer. For example, are responses more likely to display deference or caution, warmth or rigor, depth or brevity, candor or execution?

The researchers sampled just under 310,000 real conversations across the 20 most common languages, equally split between three models (Sonnet 4.6, Opus 4.6, Opus 4.7). They found that the values Claude demonstrated shifted according to the language it was addressed in, such as leaning toward greater warmth in Hindi and Arabic and toward greater rigor in English and Russian."
https://www.searchenginejournal.com/lost-in-translation-how-language-changes-which-brand-ai-recommends/591585/
#metaglossia
#metaglossia_mundus

"More Indigenous Languages Come to iPhone and iPad Keyboards in New Apple Update
Some schools are using these keyboards to help students learn these endangered languages.

Students can use the iPad and other Apple devices to study Cherokee and other Native American languages.
Apple
When Apple released iOS 27 and iPadOS 27 in September, much of the buzz centered on the update’s revamped Siri AI. However, the update also brought a handful of new indigenous languages to the device’s keyboard, allowing users to type and send messages in Blackfoot and other languages. In the US, Monday is Indigenous Peoples’ Day, a federally recognized holiday honoring Native American heritage, culture and sovereignty.

Some of the supported languages have only a couple of thousand speakers. Consider Cherokee, which has only about 1,500 fluent speakers in the entire world. Cherokee was the first Native American language Apple added to iOS back in 2010, and its inclusion has been seen by some as a welcome development.

“The technology that we utilize with Apple has allowed us to take everything that we really are trying to achieve here, which is the perpetuation and the revitalization of Cherokee language and culture, and use that same technology to make it relevant to the young people that are learning here,” Chuck Hoskin Jr., principal chief of the Cherokee Nation, said, according to Apple.

Teachers in Oklahoma are also using these tools in their classrooms to help more students engage with Native American languages. These tools can help students practice their pronunciation by allowing them to record themselves to revisit later.

“In Cherokee language, every sound is so important,” Erlinda “Daksi” Soap, a fifth-grade teacher at Oklahoma’s Cherokee Immersion School, told Apple. “One sound off, and you’re saying a completely different word.”

The Cherokee keyboard can be accessed and used across iOS devices.

There are 85 characters, originally 86, in the written Cherokee language used to represent the sounds — or syllables — in the spoken language. Cherokee artist Roy Boney Jr. worked with Apple engineers to bring the Cherokee syllabary to Apple devices.

“A big part of being Cherokee is you ensuring things like language and culture survive and continue,” he told Apple.

High-schooler Olivia Daugherty graduated from the immersion school in 2024. She said she remembers her father teaching her to count and name animals in Cherokee as a child, and that she was nervous about losing her language and culture. Now, younger generations are helping to secure the language’s future.

“I was worried about my language and culture, and the way I might lose my community, and I might lose my language,” Daugherty told Apple. “My niece, she’s a first-language speaker in our language … She’s one of the first in a very long time.”

Zachary McAuliffe
Senior writer
Edited by Gael Cooper
October 12, 2026, 3:31 pm ET
https://www.cnet.com/tech/services-and-software/indigenous-languages-iphone-ipad-keyboards-apple-update/
#metaglossia
#metaglossia_mundus

"Dubverse – Plateforme de doublage vidéo par IA, sous-titrage et synthèse vocale

Dubverse double des vidéos dans d'autres langues grâce à des voix d'intelligence artificielle, tout en générant des sous-titres et de la synthèse vocale (TTS). Vous importez un clip, choisissez une langue et une voix, et la plateforme produit une piste doublée conçue pour préserver l'émotion de l'original. Les modèles utilisés sont Neo.One et Candy.Two. Le clonage vocal permet de conserver un locuteur attitré pour une marque d'une langue à l'autre, y compris pour des tournures bilingues comme le hinglish ou le spanglish au cours d'une même session. La synchronisation labiale (lip sync) et la traduction de texte incrusté dans la vidéo font partie des fonctionnalités des forfaits supérieurs. Une API permet d'intégrer ces voix dans des applications, avec une latence annoncée de moins de 500 ms. L'abonnement Pro est affiché à environ 18 $ par mois pour 50 crédits (soit environ une minute de doublage pour quatre crédits). Le plan Supreme coûte environ 30 $. L'entreprise à l'origine du projet est Lensmatic Solutions, basée en Inde, ce qui fait des langues indiennes un point fort indéniable. Le site revendique plus de 200 voix et des millions de voix off générées chaque jour.

Forces
• Doublage, sous-titrage et synthèse vocale regroupés sous un même compte
• Des voix fidèles et adaptées aux langues indiennes comme internationales
• Clonage d'une voix de marque, au lieu de se limiter à des voix pré-existantes standard
• API dédiée aux équipes produit
• Modification du script textuel avant la génération audio finale 

Idéal pour
Les créateurs et les marques qui ont besoin d'une vidéo en hindi, en espagnol ou dans une autre langue sans avoir à passer par un studio d'enregistrement.

En bref
• Plus de 10 langues indiennes prises en charge
• Traitement par lots (batch processing) pour les séries de vidéos
• Glossaire de termes personnalisés
• Vérification humaine disponible sur l'offre Enterprise
• Approche « Made in India » garantissant une excellente restitution des accents locaux 

Potentiels problèmes
• Les crédits s'épuisent plus rapidement pour le doublage que pour les sous-titres
• Vérifiez les conditions de votre forfait en cours avant de traiter une vidéo longue
• La synchronisation labiale n'est pas parfaite sur tous les types de visages
• Une voix naturelle ne remplace pas une traduction professionnelle certifiée
• L'offre Enterprise nécessite de contacter l'équipe commerciale 

Verdict final
4.1/5 étoiles. Utilisez Dubverse lorsque la tâche consiste à dire « cette vidéo a besoin d'une autre langue cette semaine ». Écoutez un extrait de 60 secondes dans la langue cible avant de lancer le traitement de la vidéo complète. Gagnez également 1 QUA en laissant un avis sur Quasa.io !"
https://quasa.io/fr/video/dubverse-plateforme-de-doublage-video-par-ia-sous-titrage-et-synthese-vocale
#metaglossia
#metaglossia_mundus

China, the original source of vertical text, fully adopted horizontal writing after the war. Postwar Japan, by contrast, has managed to use vertical and horizontal writing skillfully depending on the purpose

"Yoroku: Let's not allow Japan to slip into a horizontal writing-only world

"It was more than 100 years after the arrival of the Dutch that the study of horizontal letters began." What Japanese scholar Sugita Genpaku (1733-1817) recorded in "The Beginning of Dutch Learning" was the Kyoho Reforms of Tokugawa Yoshimune, the eighth shogun of the Edo period (1603-1867). In 1720, the ban on importing Western books was eased, and Dutch interpreters in Nagasaki were permitted to study "horizontal letters" -- script written and read horizontally.

The reason "horizontal letters" refers to Western languages is that it rests on the premise that Japanese is written vertically. Vertical writing came from ancient China, where it was used to write characters on oracle bones and bamboo slips. On long horizontal plaques hung at temple gates and elsewhere, there are characters that appear to be horizontal writing from right to left, but apparently the correct interpretation is that they are vertical writing, one character per line, according to the book "Yokogaki tojo" authored by Makoto Yanaike and published by Iwanami Shoten Publishers.

Horizontal writing is said to have become common after the Meiji Restoration in 1868, as Japan absorbed Western culture. Before World War II, there was much horizontal writing from right to left, in keeping with vertical writing whose columns proceed in the same direction. Horizontal writing from left to right took hold after the war. Today, even textbooks are almost all written horizontally except for Japanese language texts.

"Do you have occasions in daily life to read or write characters vertically?" In the Agency for Cultural Affairs' "public opinion survey on the Japanese language," a little more than 30% answered, "No." For media that publish vertically, that is a jolting figure.

Not only newspapers but magazines and novels, too, are still mainly written vertically. The reason people can live without encountering it is probably that more and more people now rely on smartphones as their main source of information. Vertical writing does not suit internet media, which developed under the leadership of the "horizontal letters" world.

China, the original source of vertical text, fully adopted horizontal writing after the war. Postwar Japan, by contrast, has managed to use vertical and horizontal writing skillfully depending on the purpose. There is also the traditional culture of calligraphy, which presupposes vertical writing. A future of nothing but horizontal writing would be less than desirable.

("Yoroku" is a front-page column in the Mainichi Shimbun)"
October 11, 2026 (Mainichi Japan)
https://mainichi.jp/english/articles/20261009/p2a/00m/0op/005000c
#metaglossia
#metaglossia_mundus

"Anger as interpreters for foreign benefits seekers cost taxpayers' £12.9m last year
Latest official figures show translation services for benefit claimers have continued to soar year-on-year, sparking disbelief.
Tony Whitfield
Oct 10, 2026
The cost of providing translation and interpreters for foreign benefit seekers has nearly doubled in just a few years, with taxpayers picking up a £12.9million bill last year. Year-on-year the bill has increased from £2.7million in 2019-20 and providing assistance for claimants who cannot speak English increased from £6.9million in 2023-24, the DWP said. Interpretation costs made up the lion’s share, accounting for £6.2million in 2023-24. Around one in six Universal Credit claimants - 1.3 million - were foreign nationals in February.

Three fifths were EEA citizens eligible to claim benefits under the EU settlement scheme, while the remainder were eligible either because they have humanitarian, refugee, indefinite leave to remain or limited leave to remain including family reunion status.

Currently the DWP has a legal duty to provide spoken and written language services to those whose do not speak English as a first language or have hearing impairments under the Equality Act. The proportion of annual interpreter related and translation costs for deaf and hard of hearing people in 2025 is 17%, the government said.

William Yarwood, campaigns director at the TaxPayers’ Alliance, said: “Taxpayers will be alarmed that the bill for translation and interpretation at the DWP has more than doubled in just a few years. Support such as British Sign Language is clearly necessary, but ministers should separate those costs from foreign-language services, which are rising due to years of low-quality mass migration

“Clamping down on claims for foreign nationals would protect the benefits system and help cut the translation bill taxpayers are being asked to fund.”

Robert Jenrick, Reform UK MP for Newark, added: “This is yet another example of successive Tory and Labour governments’ insane waste of taxpayers’ money.

Reform UK treasury spokesman Robert Jenrick during a recent visit to Redcar (Image: Andy Commins)
“A Reform government would stop handing out benefits to foreign nationals and prioritise British citizens. Those who enter our country should speak English and be here to work, not leech off British taxpayers.”

A Government spokesperson said: “It is right that we support benefit customers to improve their chances of finding work. Help is on offer regardless of nationality, and some people who are deaf or hard of hearing rely on our translation services. Net migration continues to fall under this Government, helping to ease pressure on housing, the NHS and other public services.”

They added net migration continued to fall under the Labour Government and is now 82% below its 2023 peak, helping to ease pressure on housing, the NHS and public services.

The government has also raised the Skilled Worker visa to degree level, raised salary thresholds and raised the English language requirement to A level equivalent. More than 100 occupations have been denied access to overseas recruitment, and the Immigration Skills Charge has increased to invest in domestic training.

They stressed it provides support services to enable customers to access the benefits and other DWP services they are entitled to."

https://www.express.co.uk/news/uk/2255890/anger-interpreters-foreign-benefits-seekers
#metaglossia
#metaglossia_mundus

Along with the literal word-for-word translation, the new Smart Assist mode should help clean up spoken words and prepare them for translation.

"Google might be bringing one of the best Gboard features to Translate. A smart-assist tool has been spotted in the Google Translate app code, and it should work similarly to the popular Rambler tool in Gboard...

Along with the literal word-for-word translation, the new Smart Assist mode should help clean up spoken words and prepare them for translation.

Google created Rambler to deal with filler words, bad pronunciation, and other artifacts of spoken language. You see, when we speak, we don't sound clear and formal like a well-written article; we use a lot of filler words, pauses, filler sounds, and exclamations, and oftentimes we're just mumbling under our breath.

If you try to transcribe all these extra noises word-for-word, you end up with nonsense or, in the best-case scenario, text that's full of unwanted clutter that needs to be redacted and removed. When trying to translate such voice-to-text input into another language, things get even messier.

This could change with the new Smart Assist tool, but there might be some limitations.

Smart Assist limitations

Given that Rambler in Gboard is only supported on Pixel 11 devices, there are questions about this new feature and its compatibility.

Another interesting question is whether this Smart Assist tool will be available offline when you're not connected to the big Google data centers, or if it will require an internet connection to work properly.

It's not clear whether the models behind Smart Assist use on-device AI, whether they need to send data online, or whether they use AI in the first place. Gboard Rambler, for example, uses AI models and works differently depending on whether you're online or offline.

In online (cloud-assisted) mode Rambler is much more accurate and fast, removing complex filler words, fixing mid-sentence self-corrections, and seamlessly switching between different languages. If you're offline, the advanced features are disabled and the tool transcribes your voice with minimal and basic cleanups.

Testing this new feature, Android Authority found that Translate only appears to be planning to support Smart Assist in Conversation mode. The option is unavailable (at the moment) in both Listening and Text-only modes.

Is it a game-changer?

It could be, if executed flawlessly. I've found that sometimes using such smart assistants can result in messier situations than not using them at all.

This Smart Assist tool could, for example, start cutting out important phrases, and even though you probably won't end up in a life-threatening lost-in-translation situation, it could ruin your time abroad...

It's not clear when Google will roll out the new Smart Assist tool, or whether it will land on all Android phones or be limited to just Pixel 11 devices."
Google Translate is borrowing Gboard’s best feature — and it changes everything
Mariyan Slavov
Oct 10, 2026
https://www.phonearena.com/news/google-translate-is-borrowing-gboards-best-feature--and-it-changes-everything_id183982
#metaglossia #metaglossia_mundus

"Lost in Pace, Not in Translation: An Intern's View from Inside a Chinese AI Start-up

I arrived in June 2026 as a trained Chinese-English translator, expecting to learn about technology. I did. But the real lesson was that the hardest thing to translate between China and its Eurasian partners is not language. It is pace.

My job was to help build a distributor network for the device across the Eurasian Economic Union (EAEU). I studied competitors, analyzed marketplace data, built a database of more than a hundred potential partners and helped prepare a market-entry plan.

One of my most memorable days was at the factory, where I took part in checking the first batch of devices. Every unit was tested step by step, and every result was recorded. What struck me even more was the absence of distance. In Shenzhen, engineers, suppliers and production lines are often a short drive apart. A change discussed in the office can be tested on real units within days. Chinese hardware companies are fast not only because people work hard, but because the whole ecosystem is built for iteration.

Speed, in writing

Inside the company, decisions that would take a week of meetings elsewhere were made in a short chat thread. That speed was not chaos. Nearly all work lived in Feishu, a Chinese workplace platform: documents, tables, tasks and every conversation. Anyone could see the status of any project at any time, which made it the fastest onboarding I have ever had.

The flip side is an expectation of responsiveness. Messages arrive in the evening and on weekends, and a quick reply signals commitment. Learning where the real boundaries are is one of the first adjustments a foreigner has to make.

Equally striking was the preference for action over perfection. Managers wanted a rough answer today rather than a perfect report next month. For someone trained in a profession where every comma matters, this was uncomfortable at first. Over time I saw the logic: in a fast market, a good-enough decision made early often beats a perfect one made too late.

Hierarchy, but close at hand

Respect for seniority is real in Chinese companies, yet in a start-up the founders joined working chats, commented on drafts and asked interns for their views. Hierarchy shaped how things were said, with tact and attention to face, more than whether they could be said. The key skill for a foreigner is reading what is left unsaid. A polite "let's think about it" can mean no.

Fast meets careful
The most interesting part of my work was standing between two worlds. My Chinese colleagues thought in terms of scale, speed and product. Potential partners in the EAEU thought in terms of trust, risk and paperwork. On almost every call, the questions came in the same order: certification, warranty, exclusivity, margins.

Certification is where the gap becomes concrete. Electronics sold in the EAEU need proof of conformity marked with the EAC sign, and the document must be issued to a company registered inside the Union. A Chinese manufacturer cannot fully certify its product alone; it needs a local partner first. Add lab testing, special rules for shipping lithium-battery devices and customs clearance for samples, and it is clear why partners in the region move carefully.

Neither side was wrong, but each misread the other. A Chinese team could see a cautious partner as slow. A partner could see a fast-moving Chinese company as not serious about the long term. Much of my work was translation in the broadest sense: explaining speed to one side and caution to the other.

What I take away

Trade between China and the EAEU is growing, and more Chinese technology brands will try to enter these markets. Their success will depend less on product quality, which is often excellent, and more on people who understand both business cultures. For Chinese companies, my advice is to plan certification from day one and to treat the first local partner as a co-founder of the market, not just a buyer.

Too often, cooperation with China is discussed only in terms of trade volumes. In my experience, the real bottleneck is people who can translate between business cultures, and universities on both sides should be training them.

For students from Eurasia considering work in China, my advice is simple: go, learn the tools, learn the pace, and listen to what is left unsaid. In Shenzhen, everyone knows who does what, and by when. Between China and Eurasia, the open question is who will do the translating.

Katsiaryna Hruzheuskaya
Yazhou Bay Science and Technology City
Yazhou District, Sanya, Hainan 572025
China

Katsiaryna Hruzheuskaya is an MBA student at China Agricultural University (Sanya) and a Chinese-English translator with about five years of professional experience as a teacher, translator and executive assistant. In 2026 she completed a business development internship at a Shenzhen-based AI hardware company, helping build its partner network across the Eurasian Economic Union. She is a graduate of Minsk State Linguistic University and a winner of the 7th "Belt and Road" International Scientific and Technical Translation Competition. She writes about Chinese business culture and China-Eurasia cooperation.
10-07-2026
Press release from: Independent Author
This release was published on openPR."
https://www.openpr.com/news/4653471/lost-in-pace-not-in-translation-an-intern-s-view-from-inside
#metaglossia
#metaglossia_mundus

"Unified Speech-Text Pre-training for Speech Translation and Recognition
Yun Tang
Hongyu Gong
Ning Dong
Changhan Wang
Wei-Ning Hsu, Jiatao Gu, Alexei Baevski, Xian Li,
Abdelrahman Mohamed, Michael Auli, Juan Pino

Abstract
We describe a method to jointly pre-train speech and text in an encoder-decoder modeling framework for speech translation and recognition. The proposed method incorporates four self-supervised and supervised subtasks for cross modality learning. A self-supervised speech subtask leverages unlabelled speech data, and a (self-)supervised text to text subtask makes use of abundant text training data. Two auxiliary supervised speech tasks are included to unify speech and text modeling space. Our contribution lies in integrating linguistic information from the text corpus into the speech pre-training. Detailed analysis reveals learning interference among subtasks. Two pre-training configurations for speech translation and recognition, respectively, are presented to alleviate subtask interference. Our experiments show the proposed method can effectively fuse speech and text information into one model. It achieves between 1.7 and 2.3 BLEU improvement above the state of the art on the MuST-C speech translation dataset and comparable WERs to wav2vec 2.0 on the Librispeech speech. "
License: CC BY 4.0
arXiv:2204.05409v1 [cs.CL] 11 Apr 2022
https://arxiv.org/html/2204.05409v1
#metaglossia #metaglossia_mundus

"Who Brought Easter Eggs to Eid? Auditing LLM-Generated Cultural Translation of Math Word Problems Across Languages and Regions
Authors
Parisa Suchdev
Department of Computer Science, University of Vermont Vermont Complex Systems Institute, University of Vermont Computational Ethics Lab, University of Vermont
Juniper Lovato
Department of Computer Science, University of Vermont Vermont Complex Systems Institute, University of Vermont Computational Ethics Lab, University of Vermont Complexity Science Hub
DOI: https://doi.org/10.1609/aies.v9i4.43321
Abstract
Large language models are increasingly used to adapt math word problems for personalized learning at scale, but it remains an open question whether those adaptations are consistent across models, preserve cultural diversity at scale, and reveal which cultural entities models treat as most salient. We analyze how Claude Opus 4, GPT 4.1, and Gemini 2.5 Pro adapt 60 English math word problems into Bengali, Hindi, Punjabi (India), Urdu, Sindhi (Pakistan), Italian, and Sicilian (Italy), a language set spanning the full resource spectrum, from high-resource Italian and Hindi to under-studied Sindhi, Sicilian, and Punjabi. We annotate 6,489 entity transformations, coding whether models preserve, localize, generalize, omit, or change entities such as names, foods, and places. Models agree on transformation type in 62.5% of cases and on specific substitutions in only 33.5%, meaning model choice directly shapes which cultural world students encounter. All 21 language-model combinations show entropy collapse, with adaptation compressing rather than expanding cultural diversity. Models prioritize surface markers such as names, foods, and currencies while preserving deeper structural features such as grade-level systems that embed culturally specific assumptions. Despite prompts specifying target countries, models misattribute regional context by using Bangladeshi taka for Indian Bengali students and produce cross-cultural contamination, such as adapting egg hunts as Eid activities. Some failures are visible in individual translations. Others, including diversity collapse, systematic preference for surface markers, and consistent regional misattribution, emerge only through corpus-level analysis. The surface plausibility that makes adapted problems look correct is precisely what makes deeper failures easy to overlook."
https://ojs.aaai.org/index.php/AIES/article/view/43321
#metaglossia
#metaglossia_mundus

"Are Shared Games Building Bridges Across Borders?

A shared match gives players from different countries a clear reason to talk. Research shows how far that connection can go, and where it stops.

A shared match gives players from different countries an immediate, purposeful reason to communicate.
89% of U.S. teens who play video games play with others, according to Pew Research Center.
Repeated sessions turn recognition into familiarity, and 47% of teen players have made an online friend through a game.
A shared match can give people from different countries an immediate reason to communicate. Players do not need identical backgrounds to understand a goal, a teammate’s position, or the importance of good timing. That common focus can turn a brief interaction into a conversation with a clear purpose.

Continued access helps players remain part of the online spaces where those interactions happen. A United States Zenlesszone zero top up is one restrained example of a region-specific in-game purchase that can support participation in a shared title. The social value still comes from the communication and cooperation between the people playing.

Shared Objectives Create A Practical Social Language

Heroes in Overwatch
Gaming’s social side can be easy to miss when the visible activity is play itself. Pew Research Center found that 89% of U.S. teens who play video games play with others online or in person, while 72% said spending time with others was a reason they played. Those figures place shared play at the center of many players’ routines.

Common objectives give strangers a useful starting point. Team roles, map knowledge, visual signals, and short callouts can communicate intent without long personal introductions. Players can understand what needs attention in the moment, then build more familiar conversation over repeated sessions.

Research involving a World of Warcraft guild with 20 participants from 10 countries described online games as spaces for intercultural communication. Shared goals and cooperation helped participants work around limited language proficiency and different ideas about time. Those differences remained present, yet the game supplied a framework for adapting to them.

Repeated Play Gives Cross-Border Connection Room To Grow

League of Legends.
Players can buy digital games from official platform stores or digital marketplaces. Eneba is one option for game keys or gift cards, and its product pages can show region tags and platform information to help shoppers choose content suited to their account.

Regional availability belongs in the wider social picture because a community needs participants who can stay involved in the same game. Checking the relevant region and target platform is a straightforward part of selecting a digital purchase. It supports access while keeping attention on the relationships and routines that form during play.

Repeated sessions can turn recognition into familiarity. Guilds, squads, and community groups give people recurring contact, where they can learn another player’s communication style and establish a shared routine. In Pew’s survey, 47% of teen players said they had made an online friend because of a video game they both played, which demonstrates a real possibility rather than a universal outcome.

Cross-border communities still include different humor, customs, and expectations around teamwork. Players may simplify their language, clarify plans, or adjust to a teammate’s level of directness. An ethnographic study of an international Lords Mobile guild identified both bonding social capital within relationships and bridging social capital across cultural or social differences, alongside emotional support among members.

The evidence calls for a measured view. The World Health Organization describes research on digital play and social connection as limited and mixed, so games provide a setting where connection can develop rather than a guaranteed answer to isolation. Digital marketplaces like Eneba can offer deals on digital products and a practical route to region-specific content, while meaningful cross-border connection comes from shared play and clear communication."
By Moiz Banoori
https://exputer.com/features/shared-games-building-bridges-across-borders/
#metaglossia
#metaglossia_mundus

Can an Emoji Be Rude? A Cross-Cultural Study of Emoji Interpretation in Dig
11 Oktober 2026, 20:53 WIB

Bagikan ide kreativitasmu dalam bentuk konten di Kompasiana

Can an Emoji Be Rude? A Cross-Cultural Study of Emoji Interpretation in Digital Communication

By Kayla Putri Ramadani and Kireina Syavel Azzahra

Cross-Cultural Communication Students, Faculty of Humanities, Universitas Airlangga

Have you ever sent a crying emoji because something was funny, only to have someone think you were actually sad? I have experienced a similar misunderstanding in digital communication. An emoji I used to express laughter was interpreted as a sign of sadness. I also used a smiling emoji with a reddish face to express sarcasm, but it could be understood as a genuine smile. These experiences made me realize that emojis do not always communicate the feelings we intend to express.

Today, emojis are commonly used to express feelings that can be difficult to communicate through text. They can make digital conversations feel more like face-to-face interactions. However, people do not always interpret emojis in the same way. Cultural differences between countries, regions, and communities can influence how people understand and use them.

 

Emojis are more than just pictures that make digital conversations interesting. In face-to-face conversations, people communicate emotions through facial expressions, gestures, and tone of voice. These signals are often missing from text-based communication. Emojis can help express feelings and intentions that words alone may not convey clearly. According to Kavanagh (2016), emoticons can also convey politeness in online communication. This suggests that emojis and similar symbols serve a social purpose rather than simply decorating messages.

However, using the same emoji does not guarantee that everyone will understand it in the same way. Although emojis are used worldwide, people from different cultural backgrounds may have different habits when using and interpreting them. Guntuku et al. (2019) compared emoji usage among users from China and Japan with users from the United States, the United Kingdom, and Canada. Their study found differences in the frequency, context, and topics associated with emoji use. The researchers also identified similarities between cultures. These findings suggest that emojis are widely shared symbols, but their use can still be influenced by cultural background.

When Emojis Mean Different Things

Cultural differences can become particularly noticeable when emojis communicate meanings that are not stated directly. Sarcasm and irony are examples of this. A person may use an emoji to show that a statement should not be taken literally, while the receiver may interpret it differently.

Du (2024) examined how emojis influenced the interpretation of verbal irony among Chinese and American participants. The study found cultural differences in how participants understood irony when different emojis were used. American participants showed better understanding of irony when a winking emoji was used, while Chinese participants showed better understanding when a smiling emoji was used. This finding illustrates how a small symbol can influence the interpretation of a message.

For example, imagine someone sends the message, "Great job!" followed by a particular emoji. Depending on the situation and the relationship between the sender and receiver, the message might be interpreted as genuine praise or sarcasm. Without understanding the context, the receiver may misunderstand the sender's intention.

Emojis, Politeness, and Relationships

Emojis can also influence how people express politeness and maintain relationships. In digital conversations, an emoji can make a message seem warmer, friendlier, or less serious. Kavanagh (2016), who studied American and Japanese online blogging communities, found that both groups used emoticons to express positive politeness. However, Japanese users in the study used them more frequently than American users. This difference suggests that the use of emojis to maintain friendly communication may be connected to cultural communication patterns.

These differences do not mean that emojis are ineffective or that people should stop using them. Instead, they show that digital communication is not always as universal as it appears. People from different cultural backgrounds bring different experiences and expectations to online conversations."

Kireina Syavel Azzahra

Pelajar

https://www.kompasiana.com/kireinasyavel6777/6acb94de34777c6f3409c7a2/can-an-emoji-be-rude-a-cross-cultural-study-of-emoji-interpretation-in-dig
#metaglossia
#metaglossia_mundus

"Untangling Multiple Meanings in the English Language in the Courts

Professor's New Book Examines Ambiguities in English Language
by Alex Russell October 05, 2026
There’s a joke by British TV comedian Benny Hill that Americans usually don’t get.

Hill is dressed up as an irate mother talking to the school headmaster who says her son is always absent and now he’s boozing. Hill insists that there are 40 pubs in town, and Johnny hasn't been in one of them. The headmaster says, “Oh, yes, which one's that?”

The humor comes from a very subtle ambiguity in what both sides mean by Johnny being in “not in one” pub. Hill means Johnny hasn’t been in any pubs, but the headmaster is implying that Johnny has more likely been in all of them.

But not all ambiguity in language is funny. In a court of law, the way ambiguity is interpreted can have repercussions measured in the billions of dollars.

“What frequently happens?” said John Hawkins, a Distinguished Professor of linguistics at UC Davis. “One side picks the one meaning and the other side picks the other. What you need is a linguist to come in and say, ‘Well, how does disambiguation normally work?’”

Hawkins is just such a linguist, and he has served as an expert witness in numerous high-profile court cases that hinge on disambiguation, which is the process of removing the multiple meanings contained in a passage of text.

His new book, The English Language on Trial (Cambridge University Press, 2026), untangles some of the fundamental linguistic issues underpinning cases involving high-profile clients including Hollywood celebrities, major league sports teams and some of the world’s largest corporations. Each of them came down to some of the built-in ambiguities or other subtleties in the English language itself.

“People wouldn't normally disagree on these things if it was outside the courtroom, and these are systematic and there's many of them,” said Hawkins. “I've written many reports over the years that have tried to explain to lawyers and to the courts what the linguistic issue is, and what the proper judgment should be.”

Ambiguity in the English Language
The English language has many more ambiguities than many other languages, said Hawkins, and for specific reasons. One is that over the past millennium, English has undergone incredible simplification in its word forms compared to its predecessor language Old English. These changes were driven by contact between the English and populations from Denmark and other Vikings, as well as by the 1066 conquest by the French-speaking Normans.

“English has been exposed to a battering from languages and contact over its history, more than other similar languages have,” said Hawkins.

Ambiguity is rampant in English partly because so many of its verbs are used in ways they are not in other languages. For example, in English, you can break both a leg and a tennis racket string, but other languages use different words depending on what’s broken. Spanish has both romper, used in the general sense for causing damage, and quebrar, used for brittle things that shatter or for bankruptcy.

Another source of ambiguity are words with multiple meanings. Port is a type of wine as well as a place to dock a boat or to plug in a phone charger. Knowing the context of the sentence eliminates the alternative meanings.

In court cases, the word “substantial” comes up often, and it’s a word that almost could not be more ambiguous. The sun is substantially larger than the Earth, which compares the size of both celestial bodies, but your homework can also be substantially finished when most, but not all of it, is.

But what if a legal contract states that a billboard must be "substantially below" the top of a building’s roof? Does all of it have to be below? How far below? Here, the context must make clear what was intended.

Taking ambiguity in the English language to court
In total, Hawkins has served as an expert witness in over 50 cases. In his book, he details what many of the fundamental linguistic issues are and what the proper judgment should be.

In some of the cases Hawkins has worked on, he had to look closely at the broader context around the language under dispute. What other bits of language are relevant? How will average readers understand it? Sometimes that context involves how a word is used across society.

In a case involving the company Salesforce, Hawkins conducted an extensive analysis over the spelling of the company name that would be used to determine whether it could be used as the company trademark.

“When you have a trademark, you have to protect it,” said Hawkins.

Sometimes companies don’t succeed. Xerox and aspirin are two examples of “genericide,” in which a brand name becomes generic through common use.

At issue in the Salesforce case was how different is that one-word spelling from the general use of “sales force” as two separate words to describe personnel at any company who are responsible for sales.

Hawkins conducted a comprehensive database search to find out. He counted the instances in which the one-word version identified either the company or a corporate sales team and vice versa. He found that overwhelmingly the single-word version, “Salesforce,” was used in the context of the company itself.

Hawkins was also an expert witness in Apple’s 2013 trademark dispute against Amazon over the use of the name “App Store.” Apple claimed trademark for the name, arguing it had been the first to use it.

One of the lawyers Apple hired was Robert Leonard, a well-regarded linguist, who came to one of Hawkins’ depositions and supplied questions to Apple’s legal team. Leonard leads the Forensic Linguistics Justice Project at Hofstra University that reanalyzes cases in which language evidence played a crucial role in the defendant's conviction and sentence of death.

In the end, Hawkins’ analysis showed that “app store” truly was used generically. Even Apple CEO Tim Cook had used the term generically in an earnings call. In the U.S., Apple went on to withdraw its claim before trial.

“If you've got a phone from a manufacturer other than Apple and you have an ‘app store,’ it’s because of me,” said Hawkins."
https://www.ucdavis.edu/magazine/untangling-multiple-meanings-english-language-courts
#metaglossia
#metaglossia_mundus

"Tomedes Expands Its AI Translation Platform for Professional Translators With CAT Tool File Support
Translators can now upload XLIFF and MQXLIFF files from MemoQ, Trados, and Phrase to MachineTranslation.com and import the AI-translated file straight back into their CAT tool.

BEAVERTON, Ore. - Oct. 9, 2026 - PRLog -- Tomedes, a professional translation company, has extended its AI translation platform, MachineTranslation.com, to work directly with XLIFF and MQXLIFF files. Translators using computer-assisted translation (CAT) tools can now upload project files and receive the translated file back in its original format.

CAT tools such as MemoQ, Trados, Phrase, Wordfast, and OmegaT exchange project content through XLIFF, a file that holds source text, target segments, and project metadata together. Previously, using MachineTranslation.com with this content meant copying the text into a separate document and transferring the results back into the project by hand.

Workflow
Export an XLIFF or MQXLIFF file from the CAT tool.
Upload it to MachineTranslation.com and choose the languages.
Download the translated file in the same format.
Import it back into the CAT tool for review.

During processing, the platform sends each segment to multiple AI models, compares their translations, and keeps the version most of them agree on.

"Translators have spent years setting up their CAT tools, and that setup is where their work happens," said Ofer Tirosh, CEO of Tomedes. "XLIFF and MQXLIFF support lets them bring MachineTranslation.com into a project without leaving that environment."

Comparing output across models

Many AI features in CAT tools rely on a single model that offers one suggestion per segment. MachineTranslation.com translates the same text with 22 AI models and selects the version most of them produce, setting aside translations that differ from the majority.

"XLIFF support was one of the most frequent requests from our professional users," said Shashank Jain, Tech Lead at Tomedes. "Handling the format on both upload and download takes manual conversion out of the process."

XLIFF and MQXLIFF support is now live on https://www.machinetranslation.com/, which covers more than 330 languages.

About MachineTranslation.com

MachineTranslation.com is an AI translation platform built by https://www.tomedes.com/, a professional translation company with nearly 20 years in the language industry. It has AI models such as ChatGPT, Claude, Gemini, DeepL, and Google translate the same text separately, then shows the translation most of them agree on.

Media Contact
Rachelle Garcia
rachelle@tomedes.net"
https://www.prlog.org/13175725-tomedes-expands-its-ai-translation-platform-for-professional-translators-with-cat-tool-file-support.html
#metaglossia
#metaglossia_mundus

"The growing use of artificial intelligence in online search is changing how consumers discover information about businesses, products, and services. Rather than relying exclusively on traditional search engine results, consumers increasingly encounter AI-generated answers that summarize information gathered from websites and other digital sources.

For businesses, this development introduces additional considerations regarding website content, organization, and the accuracy of publicly available information.

Rhino Web Studios, a New Orleans-based website development and digital marketing company, is examining how website content contributes to business visibility within AI-generated search responses.

Traditional search engines generally present lists of websites based on relevance, authority, and other ranking factors. AI-powered search systems may instead generate conversational responses that combine information from multiple sources. Depending on the platform and search query, responses may include citations, business descriptions, recommendations, or links to supporting websites.

Technical considerations remain important. Pages blocked from crawling, dependent on inaccessible content, or affected by indexing problems may be difficult for search systems to retrieve.

Structured data markup can provide machine-readable information about certain website elements, including business details, articles, and frequently asked questions. However, structured data alone does not establish eligibility for inclusion in AI-generated answers.

Answering Common Questions Through Website Content

Informational articles represent another component of website content development.

Businesses frequently receive questions concerning service procedures, materials, maintenance requirements, pricing considerations, and industry terminology.

Publishing factual explanations of these subjects can establish additional context surrounding a company’s activities.

A heating and air conditioning contractor, for example, might publish articles explaining refrigerant leaks, airflow restrictions, seasonal maintenance, or common causes of equipment failure.

Such information can address consumer questions while creating relevant material that search systems may retrieve when processing related inquiries.

Thomas noted that the usefulness of informational content depends partly on its accuracy and specificity.

“An article should provide meaningful information about the subject being discussed,” Thomas said. “Simply repeating keywords or publishing large amounts of generic material does not necessarily help an AI system understand a business. Accurate explanations, relevant details, and clearly identified subjects provide more useful context.”

Business Information Consistency Remains Important

Information consistency across digital platforms is another consideration.

Business names, physical addresses, telephone numbers, operating hours, and service descriptions may appear on company websites, business directories, mapping platforms, and industry listings.

Conflicting information can create uncertainty for consumers and automated systems attempting to identify a particular business.

Maintaining accurate business details across relevant platforms can support a more consistent digital identity.

Company websites can also provide information about professional experience, geographic service areas, licensing where applicable, and the individuals responsible for business operations.

These details may help establish context, although AI search systems apply different methods when evaluating sources and generating responses.

AI Search Does Not Replace Traditional Search Optimization

The emergence of AI-generated answers does not eliminate the role of traditional search engine optimization.

Search engines continue to use established crawling, indexing, and ranking processes. Many AI-powered search experiences also depend on search infrastructure to retrieve supporting information.

Consequently, website accessibility, technical performance, relevant content, and logical information architecture remain important considerations.

Differences between AI platforms also make visibility difficult to predict. A business cited by one system may not appear in another system’s response to a similar question.

AI-generated answers can also change over time as underlying models, retrieval systems, and available information evolve.

“Businesses should understand that appearing in an AI-generated answer is not something that can be guaranteed,” Thomas said. “The practical focus should remain on maintaining a website that accurately explains the business, provides useful information, and makes that information accessible to search technologies.”

An Evolving Approach to Online Information

As AI-powered search continues to develop, website content is becoming relevant to multiple forms of online information discovery.

Traditional search results, AI-generated summaries, conversational search interfaces, and other digital platforms may interpret and present business information differently.

For businesses, the underlying consideration remains the same: publicly accessible information should accurately represent the organization and provide meaningful explanations of its activities.

The continuing development of AI search technologies is likely to influence how businesses organize, maintain, and evaluate website content in the years ahead.

About Rhino Web Studios

Rhino Web Studios is a New Orleans, Louisiana-based website development and digital marketing company providing website design, search engine optimization, website hosting, content development, and related digital services. The company works with businesses across multiple industries to develop and maintain their online presence.

Morgan Thomas
Rhino Digital, LLC
+1 504-875-5036"
https://www.courier-journal.com/press-release/story/239733/website-content-plays-a-growing-role-in-how-businesses-appear-in-ai-generated-search-answers/
#metaglossia #metaglossia_mundus

«Allô ? Oui bonjour, j’ai un texte IA à vous faire reprendre.» Dans les agences de traduction, les propositions de ce genre sont devenues monnaie courante

«Je me suis senti insulté» : les traducteurs excédés d’être réduits à de simples «toiletteurs d’IA»

"De plus en plus de contrats ne proposent que de «nettoyer» une traduction générée par l’intelligence artificielle. Bafouant le savoir-faire et la dignité du métier, s’insurgent des traducteurs auprès du Figaro.

«Allô ? Oui bonjour, j’ai un texte IA à vous faire reprendre.» Dans les agences de traduction, les propositions de ce genre sont devenues monnaie courante. Au lieu de traduire un document «from scratch», soit «de zéro», les traducteurs reçoivent une version produite par l’intelligence artificielle générative, et doivent la «nettoyer», comprenez la rendre exploitable. Une pratique qui dénature le métier, en retirant tout le plaisir attelé à la traduction et posant de nombreux problèmes. Petit à petit, les rôles s’inversent : les traducteurs sont-ils en passe de devenir des… assistants d’IA ?

Si le terme «post-édition» s’est largement répandu pour qualifier cette nouvelle manière de travailler, les traducteurs lui préfèrent les termes «nettoyage» ou «toilettage» de «sortie machine», comme nous l’indique Laura Hurot, traductrice de l’allemand et membre du collectif IA-lerte générale. Son confrère Cyril Laumonier, président de l’Association des traducteurs littéraires de France…"
Par Romain Ferrier
Publié le 11/10/2026 à 08
https://www.lefigaro.fr/langue-francaise/actu-des-mots/je-me-suis-senti-insulte-les-traducteurs-excedes-d-etre-reduits-a-de-simples-toiletteurs-d-ia-20261011
#metaglossia
#metaglossia_mundus