 Your new post is loading...
|
Scooped by
Charles Tiayon
November 25, 2011 12:05 PM
|
Si vous n’avez pas l’habitude de recourir à un traducteur professionnel, entendre votre prestataire évoquer une relecture externe, par une tierce personne spécialisée, peut vous surprendre, voire éveiller votre méfiance. « Comment ? Voilà quelqu’un qui se vante d’être professionnel et spécialiste de mon métier, qui m’annonce un prix en rapport, et qui a si peu confiance dans son travail qu’il le fait vérifier et corriger par un autre ? » Loin de vous alarmer, cette démarche doit au contraire vous rassurer, car elle témoigne du sérieux et du professionnalisme de votre interlocuteur.
Researchers across Africa, Asia and the Middle East are building their own language models designed for local tongues, cultural nuance and digital independence
"In a high-stakes artificial intelligence race between the United States and China, an equally transformative movement is taking shape elsewhere. From Cape Town to Bangalore, from Cairo to Riyadh, researchers, engineers and public institutions are building homegrown AI systems, models that speak not just in local languages, but with regional insight and cultural depth.
The dominant narrative in AI, particularly since the early 2020s, has focused on a handful of US-based companies like OpenAI with GPT, Google with Gemini, Meta’s LLaMa, Anthropic’s Claude. They vie to build ever larger and more capable models. Earlier in 2025, China’s DeepSeek, a Hangzhou-based startup, added a new twist by releasing large language models (LLMs) that rival their American counterparts, with a smaller computational demand. But increasingly, researchers across the Global South are challenging the notion that technological leadership in AI is the exclusive domain of these two superpowers.
Instead, scientists and institutions in countries like India, South Africa, Egypt and Saudi Arabia are rethinking the very premise of generative AI. Their focus is not on scaling up, but on scaling right, building models that work for local users, in their languages, and within their social and economic realities.
“How do we make sure that the entire planet benefits from AI?” asks Benjamin Rosman, a professor at the University of the Witwatersrand and a lead developer of InkubaLM, a generative model trained on five African languages. “I want more and more voices to be in the conversation”.
Beyond English, beyond Silicon Valley
Large language models work by training on massive troves of online text. While the latest versions of GPT, Gemini or LLaMa boast multilingual capabilities, the overwhelming presence of English-language material and Western cultural contexts in these datasets skews their outputs. For speakers of Hindi, Arabic, Swahili, Xhosa and countless other languages, that means AI systems may not only stumble over grammar and syntax, they can also miss the point entirely.
“In Indian languages, large models trained on English data just don’t perform well,” says Janki Nawale, a linguist at AI4Bharat, a lab at the Indian Institute of Technology Madras. “There are cultural nuances, dialectal variations, and even non-standard scripts that make translation and understanding difficult.” Nawale’s team builds supervised datasets and evaluation benchmarks for what specialists call “low resource” languages, those that lack robust digital corpora for machine learning.
It’s not just a question of grammar or vocabulary. “The meaning often lies in the implication,” says Vukosi Marivate, a professor of computer science at the University of Pretoria, in South Africa. “In isiXhosa, the words are one thing but what’s being implied is what really matters.” Marivate co-leads Masakhane NLP, a pan-African collective of AI researchers that recently developed AFROBENCH, a rigorous benchmark for evaluating how well large language models perform on 64 African languages across 15 tasks. The results, published in a preprint in March, revealed major gaps in performance between English and nearly all African languages, especially with open-source models.
Similar concerns arise in the Arabic-speaking world. “If English dominates the training process, the answers will be filtered through a Western lens rather than an Arab one,” says Mekki Habib, a robotics professor at the American University in Cairo. A 2024 preprint from the Tunisian AI firm Clusterlab finds that many multilingual models fail to capture Arabic’s syntactic complexity or cultural frames of reference, particularly in dialect-rich contexts.
Governments step in
For many countries in the Global South, the stakes are geopolitical as well as linguistic. Dependence on Western or Chinese AI infrastructure could mean diminished sovereignty over information, technology, and even national narratives. In response, governments are pouring resources into creating their own models.
Saudi Arabia’s national AI authority, SDAIA, has built ‘ALLaM,’ an Arabic-first model based on Meta’s LLaMa-2, enriched with more than 540 billion Arabic tokens. The United Arab Emirates has backed several initiatives, including ‘Jais,’ an open-source Arabic-English model built by MBZUAI in collaboration with US chipmaker Cerebras Systems and the Abu Dhabi firm Inception. Another UAE-backed project, Noor, focuses on educational and Islamic applications.
In Qatar, researchers at Hamad Bin Khalifa University, and the Qatar Computing Research Institute, have developed the Fanar platform and its LLMs Fanar Star and Fanar Prime. Trained on a trillion tokens of Arabic, English, and code, Fanar’s tokenization approach is specifically engineered to reflect Arabic’s rich morphology and syntax.
India has emerged as a major hub for AI localization. In 2024, the government launched BharatGen, a public-private initiative funded with 235 crore (€26 million) initiative aimed at building foundation models attuned to India’s vast linguistic and cultural diversity. The project is led by the Indian Institute of Technology in Bombay and also involves its sister organizations in Hyderabad, Mandi, Kanpur, Indore, and Madras. The programme’s first product, e-vikrAI, can generate product descriptions and pricing suggestions from images in various Indic languages. Startups like Ola-backed Krutrim and CoRover’s BharatGPT have jumped in, while Google’s Indian lab unveiled MuRIL, a language model trained exclusively on Indian languages. The Indian governments’ AI Mission has received more than180 proposals from local researchers and startups to build national-scale AI infrastructure and large language models, and the Bengaluru-based company, AI Sarvam, has been selected to build India’s first ‘sovereign’ LLM, expected to be fluent in various Indian languages.
In Africa, much of the energy comes from the ground up. Masakhane NLP and Deep Learning Indaba, a pan-African academic movement, have created a decentralized research culture across the continent. One notable offshoot, Johannesburg-based Lelapa AI, launched InkubaLM in September 2024. It’s a ‘small language model’ (SLM) focused on five African languages with broad reach: Swahili, Hausa, Yoruba, isiZulu and isiXhosa.
“With only 0.4 billion parameters, it performs comparably to much larger models,” says Rosman. The model’s compact size and efficiency are designed to meet Africa’s infrastructure constraints while serving real-world applications. Another African model is UlizaLlama, a 7-billion parameter model developed by the Kenyan foundation Jacaranda Health, to support new and expectant mothers with AI-driven support in Swahili, Hausa, Yoruba, Xhosa, and Zulu.
India’s research scene is similarly vibrant. The AI4Bharat laboratory at IIT Madras has just released IndicTrans2, that supports translation across all 22 scheduled Indian languages. Sarvam AI, another startup, released its first LLM last year to support 10 major Indian languages. And KissanAI, co-founded by Pratik Desai, develops generative AI tools to deliver agricultural advice to farmers in their native languages.
The data dilemma
Yet building LLMs for underrepresented languages poses enormous challenges. Chief among them is data scarcity. “Even Hindi datasets are tiny compared to English,” says Tapas Kumar Mishra, a professor at the National Institute of Technology, Rourkela in eastern India. “So, training models from scratch is unlikely to match English-based models in performance.”
Rosman agrees. “The big-data paradigm doesn’t work for African languages. We simply don’t have the volume.” His team is pioneering alternative approaches like the Esethu Framework, a protocol for ethically collecting speech datasets from native speakers and redistributing revenue back to further development of AI tools for under-resourced languages. The project’s pilot used read speech from isiXhosa speakers, complete with metadata, to build voice-based applications.
In Arab nations, similar work is underway. Clusterlab’s 101 Billion Arabic Words Dataset is the largest of its kind, meticulously extracted and cleaned from the web to support Arabic-first model training.
The cost of staying local
But for all the innovation, practical obstacles remain. “The return on investment is low,” says KissanAI’s Desai. “The market for regional language models is big, but those with purchasing power still work in English.” And while Western tech companies attract the best minds globally, including many Indian and African scientists, researchers at home often face limited funding, patchy computing infrastructure, and unclear legal frameworks around data and privacy.
“There’s still a lack of sustainable funding, a shortage of specialists, and insufficient integration with educational or public systems,” warns Habib, the Cairo-based professor. “All of this has to change.”
A different vision for AI
Despite the hurdles, what’s emerging is a distinct vision for AI in the Global South – one that favours practical impact over prestige, and community ownership over corporate secrecy.
“There’s more emphasis here on solving real problems for real people,” says Nawale of AI4Bharat. Rather than chasing benchmark scores, researchers are aiming for relevance: tools for farmers, students, and small business owners.
And openness matters. “Some companies claim to be open-source, but they only release the model weights, not the data,” Marivate says. “With InkubaLM, we release both. We want others to build on what we’ve done, to do it better.”
In a global contest often measured in teraflops and tokens, these efforts may seem modest. But for the billions who speak the world’s less-resourced languages, they represent a future in which AI doesn’t just speak to them, but with them."
Sibusiso Biyela, Amr Rageh and Shakoor Rather
20 May 2025
https://www.natureasia.com/en/nmiddleeast/article/10.1038/nmiddleeast.2025.65
#metaglossia_mundus
Ram Mohan advocates for AI approaches that help preserve endangered languages and promote digital inclusion.
"How Multilingual AI Can Protect Language and Improve Global Technology
RAM MOHAN / JUN 5, 2025
Ram Mohan is Chairman and Founder of the Coalition on Digital Impact (CODI), an independent, global coalition founded to empower global communities to access and navigate the Internet in their native languages.
The modern ability to digitize medical records, financial transactions, and cultural documents means that important items can be recorded for posterity. But when it comes to language, technology’s rapid advancement could lead to the extinction of global languages.
For example, in southwestern Ethiopia, the Ongota language has virtually no digital presence, lacking a standardized script, keyboard support, app localization, digitized content, or representation in AI language models. Ongota speakers—and their knowledge systems—are effectively invisible in the digital world, and the language will likely die with the handful of elderly people who still speak it.
Current estimates predict that over half of the world’s languages will become extinct within 75 years, with some estimates suggesting that one language dies every two weeks. AI has created new possibilities for communication, with instant translation services and voice assistants. But while AI has the potential to bridge linguistic divides, it also poses a threat to the very languages it seeks to translate and understand.
As governments and private companies race to develop AI, large language models (LLMs) like ChatGPT are being trained with a handful of dominant, data-rich languages, such as English, Spanish, and Mandarin. Ignoring countless languages and the communities they represent has created an AI landscape that is heavily skewed toward the Western world. As AI becomes more integrated into sectors like education, healthcare, and governance, the prevalence of a handful of languages in these systems risks leading to linguistic homogenization.
Harnessing AI to Tackle Today’s Challenges
The digital access movement encompasses all efforts to address economic, cultural, and technical disparities that limit access to the internet and the opportunities it creates. But language has been overlooked, despite being a key factor in forming meaningful connections online.
OUR CONTENT DELIVERED TO YOUR INBOX.
Join our newsletter on issues and ideas at the intersection of tech & democracy
Enter email address
Subscribe
Language preservation must move beyond oral traditions and become part of the digital age. We must work to design AI systems that embrace linguistic diversity and integrate multilingualism into the internet infrastructure to avoid erasure and improve our digital world. AI systems can be part of the solution, helping to document and revive endangered languages through text, speech recognition, and translation tools. So, how can we build a world where everyone can navigate the internet in their own language?
Internet pioneer Vint Cerf says, “Meaningful spoken and written linguistic access is necessary to realize the Internet's full potential. Absent effective action, many communities around the world will be wrongly excluded. This cannot be an acceptable outcome."
The newly launched Coalition on Digital Impact (CODI), an independent, global alliance of internet leaders and like-minded organizations, is rising to meet this challenge. Through education, advocacy, and awareness efforts, CODI and its members are working to reduce barriers to internet access, build a more secure infrastructure for the future of the web, and advance policies that promote equitable access to the digital world.
Speakers of minority languages are largely unable to reap the benefits of AI. A child in a remote village may struggle to find educational content in their native tongue or may face challenges in using voice-activated devices or translation services. Without digital tools, the practical advantages of maintaining a language diminish, hastening its extinction.
Universal Acceptance: A Small Change With Outsized Benefits
For AI to be a force for good in the digital age, we need to foster a culture of multilingualism and 'Universal Acceptance' online. One solution is for organizations to implement Universal Acceptance (UA) protocols, the principle that all valid domain names, email addresses, and internet protocols should be accepted and recognized by all internet systems, regardless of language, script, or other characteristics. Non-Latin scripts such as Arabic, Chinese, and Cyrillic are increasingly being used for domain names, and UA ensures that systems can handle these characters and scripts. UA-compliance can be achieved through simple code edits or features like voice-to-text or live captions ― small changes with far-reaching benefits.
Designing products and services for a wide range of users, without the need for adaptations or specialized features, is essential for global market participation. A 2017 study found that the Universal Acceptance of internet domain names could be a USD $9.8 billion+ opportunity. UA represents the future of the web ― one that is backed by a more democratic, secure infrastructure ― and an enormous economic opportunity. But this future will depend on global collaboration and investment.
A truly multilingual internet requires addressing language barriers, literacy challenges, cultural sensitivities, and social stigmas that impede technology adoption. AI will play a decisive role in the future of global languages – it will either preserve them or accelerate their extinction. By developing AI systems that reflect linguistic diversity and building multilingualism and Universal Acceptance into internet systems, we can create a digital world where every linguistic group has a place. The time to act is now."
Ram Mohan
https://www.techpolicy.press/how-multilingual-ai-can-protect-language-and-improve-global-technology/
#metaglossia_mundus
"Nonbinary Hebrew transforms the language for everyone
As a living language, Hebrew is constantly evolving to adapt to the needs of its speakers. That must include nonbinary people too.
How do people who go by they/them pronouns in English refer to themselves in Hebrew? What do you call the Jewish rite of passage ceremony for a nonbinary tween? What plural words should we use for a group of people who each have a different gender?
I was faced with these wonderfully productive dilemmas when I studied Hebrew in college eight years ago. I had come out a year earlier as nonbinary, which means for me that my gender does not fit neatly into the boxes of man or woman. I asked Eyal Rivlin, my Hebrew professor at University of Colorado at Boulder, about the established conventions for nonbinary people. After he and I did research by asking friends, family and colleagues and examining literature, we realized there was no comprehensive system for speaking Hebrew without using the masculine or feminine.
We decided to experiment and created the Nonbinary Hebrew Project in 2018.
Hebrew is both ancient and modern. As a living language, it is constantly changing, evolving and growing to adapt to the needs of its speakers. It is used daily for conversation, prayer, ritual and study. As such, it is critical that everyone who needs or wants to use Hebrew can do so in a way that is affirming.
One of the aspects of Hebrew that distinguishes it from English is its use of what linguists call grammatical gender. English has some familiar uses of it, such as the personal pronouns she/her and he/him. But in many languages, including Hebrew, almost all parts of speech are gendered, including verbs, nouns and adjectives. This grammatical gender is often chosen based on the gender of the speaker or the subject of the sentence.
Some queer communities in Israel use “lashon me’orevet,” or “language that crosses over,” in which they intentionally challenge this convention by switching grammatical gender mid-sentence or in every other sentence for the same speaker or subject. This is affirming for many people — but I sought to create a third option for people like myself for whom the grammatical masculine or feminine did not entirely affirm my identity.
The system that Eyal and I created is intuitive to Hebrew speakers because it creates a third parallel system of grammatical gender for use alongside the masculine and feminine options.
In traditional Hebrew, for example, grammatically masculine words are usually considered the default, while grammatically feminine words often end with an additional “-et” or “-ah.” In our new, more expansive option, many singular nouns, verbs and adjectives end with “-eh” to distinguish them from the masculine or feminine.
In another example from traditional Hebrew, grammatically masculine plural words end in “-im,” with groups referred to with the pronouns “hem” or “atem,” while feminine plural words end in “-ot,” with groups referred to with “hen” or “aten.” In our new system, plural words can use the ending “-imot” or “-emen,” both of which combine existing plural endings and were already used by some people before our project.
Since the Nonbinary Hebrew Project’s creation, many people across the world have applied our system for their own uses, such as Yizkor (memorial) prayers, baby-naming ceremonies and wedding blessings. There are other innovations as well, such as using “bet mitzvot,” instead of a bar or bat mitzvah. I can now start the day with the prayer “Modet Ani” (rather than “modeh” or “modah”). And when I’m called to the Torah, I can receive a blessing that truly honors me.
Another application I have been excited to see is the use of our system to refer to the Divine without using the masculine or feminine.
There is a rich history in Hebrew texts of gender being used in playful ways, for both people and the Divine. Many names for the Divine, such as Rock or Fountain of Life, are already beyond traditional notions of binary gender.
The Nonbinary Hebrew Project opens up possibilities for radical joy, euphoria and recognition of the other as a sibling rather than a stranger. We connect with one another through shared language, and our project aims to provide another tool for weaving communities closer together.
If you’re interested in learning more, come to one of our workshops this Friday evening, June 6, at Congregation Sha’ar Zahav in San Francisco or stay tuned for our virtual workshops by checking the calendar at nonbinaryhebrew.com. You can find applied uses of the system on the website, as well as grammar charts, podcasts and news articles.
This Pride Month, and all year long, let’s find joy together by using language to uplift one another and honor each other’s light."
BY LIOR GROSS
JUNE 5, 2025
https://jweekly.com/2025/06/05/nonbinary-hebrew-transforms-the-language-for-everyone/
#metaglossia_mundus
"Amazon Lex extends custom vocabulary feature to additional languages
Posted on: Jun 4, 2025
Amazon Lex now extends custom vocabulary support to multiple languages, including Chinese, Japanese, Korean, Portuguese, Catalan, French, German, and Spanish locales. This enhancement enables you to improve speech recognition accuracy for domain-specific terminology, proper nouns, and rare words across a wider range of languages, creating more natural and accurate conversational experiences. With custom vocabulary, you can provide Amazon Lex with specific phrases that should be recognized during audio conversations, even when the spoken audio might be ambiguous. For example, you can ensure technical terms like "Cognito" or industry-specific vocabulary like "solvency" are correctly transcribed during bot interactions, providing consistent speech recognition capabilities that work both for intent recognition and improving slot value elicitation.
This feature is now available in all AWS Regions where Amazon Lex operates for the supported languages."
https://aws.amazon.com/about-aws/whats-new/2025/06/amazon-lex-custom-vocabulary-additional-languages/
#metaglossia_mundus
Dans un futur proche, la proportion de données en langage naturel sur le Web pourrait diminuer au point d’être éclipsée par des textes générés par l’Intelligence artificielle.
Les agents conversationnels tels que ChatGPT facilitent parfois notre quotidien en prenant en charge des tâches rébarbatives. Mais ces robots intelligents ont un coût. Leur bilan carbone et hydrique désastreux est désormais bien connu. Un autre aspect très préoccupant l’est moins : l’intelligence artificielle pollue les écrits et perturbe l’écosystème langagier, au risque de compliquer l’étude du langage.
Une étude publiée en 2023 révèle que l’utilisation de l’intelligence artificielle (IA) dans les publications scientifiques a augmenté significativement depuis le lancement de ChatGPT (version 3.5). Ce phénomène dépasse le cadre académique et imprègne une part substantielle des contenus numériques, notamment l’encyclopédie participative Wikipedia ou la plate-forme éditoriale états-unienne Medium.
Le problème réside d’abord dans le fait que ces textes sont parfois inexacts, car l’IA a tendance à inventer des réponses lorsqu’elles ne figurent pas dans sa base d’entraînement. Il réside aussi dans leur style impersonnel et uniformisé.
La contamination textuelle par l’IA menace les espaces numériques où la production de contenu est massive et peu régulée (réseaux sociaux, forums en ligne, plates-formes de commerce…). Les avis clients, les articles de blog, les travaux d’étudiants, les cours d’enseignants sont également des terrains privilégiés où l’IA peut discrètement infiltrer des contenus générés et finalement publiés.
La tendance est telle qu’on est en droit de parler de pollution textuelle. Les linguistes ont de bonnes raisons de s’en inquiéter. Dans un futur proche, la proportion de données en langues naturelles sur le Web pourrait diminuer au point d’être éclipsée par des textes générés par l’IA. Une telle contamination faussera les analyses linguistiques et conduira à des représentations biaisées des usages réels du langage humain. Au mieux, elle ajoutera une couche de complexité supplémentaire à la composition des échantillons linguistiques que les linguistes devront démêler.
Read more: L’IA au travail : un gain de confort qui pourrait vous coûter cher
Quel impact sur la langue ?
Cette contamination n’est pas immédiatement détectable pour l’œil peu entraîné. Avec l’habitude, cependant, on se rend compte que la langue de ChatGPT est truffée de tics de langage révélateurs de son origine algorithmique. Il abuse aussi bien d’adjectifs emphatiques, tels que « crucial », « essentiel », « important » ou « fascinant », que d’expressions vagues (« de nombreux… », « généralement… »), et répond très souvent par des listes à puces ou numérotées. Il est possible d’influer sur le style de l’agent conversationnel, mais c’est le comportement par défaut qui prévaut dans la plupart des usages.
Un article de Forbes publié en décembre 2024 met en lumière l’impact de l’IA générative sur notre vocabulaire et les risques pour la diversité linguistique. Parce qu’elle n’emploie que peu d’expressions locales et d’idiomes régionaux, l’IA favoriserait l’homogénéisation de la langue. Si vous demandez à un modèle d’IA d’écrire un texte en anglais, le vocabulaire employé sera probablement plus proche d’un anglais global standard et évitera des expressions typiques des différentes régions anglophones.
L’IA pourrait aussi simplifier considérablement le vocabulaire humain, en privilégiant certains mots au détriment d’autres, ce qui conduirait notamment à une simplification progressive de la syntaxe et de la grammaire. Comptez le nombre d’occurrences des adjectifs « nuancé » et « complexe » dans les sorties de l’agent conversationnel et comparez ce chiffre à votre propre usage pour vous en rendre compte.
Du lundi au vendredi + le dimanche, recevez gratuitement les analyses et décryptages de nos experts pour un autre regard sur l’actualité. Abonnez-vous dès aujourd’hui !
Ce qui inquiète les linguistes
La linguistique étudie le langage comme faculté qui sous-tend l’acquisition et l’usage des langues. En analysant les occurrences linguistiques dans les langues naturelles, les chercheurs tentent de comprendre le fonctionnement des langues, qu’il s’agisse de ce qui les distingue, de ce qui les unit ou de ce qui en fait des créations humaines. La linguistique de corpus se donne pour tâche de collecter d’importants corpus textuels pour modéliser l’émergence et l’évolution des phénomènes lexicaux et grammaticaux.
Les théories linguistiques s’appuient sur des productions de locuteurs natifs, c’est-à-dire de personnes qui ont acquis une langue depuis leur enfance et la maîtrisent intuitivement. Des échantillons de ces productions sont rassemblés dans des bases de données appelées corpus. L’IA menace aujourd’hui la constitution et l’exploitation de ces ressources indispensables.
Pour le français, des bases comme Frantext (qui rassemble plus de 5 000 textes littéraires) ou le French Treebank (qui contient plus de 21 500 phrases minutieusement analysées) offrent des contenus soigneusement vérifiés. Cependant, la situation est préoccupante pour les corpus collectant automatiquement des textes en ligne. Ces bases, comme frTenTen ou frWaC, qui aspirent continuellement le contenu du Web francophone, risquent d’être contaminées par des textes générés par l’IA. À terme, les écrits authentiquement humains pourraient devenir minoritaires.
Les corpus linguistiques sont traditionnellement constitués de productions spontanées où les locuteurs ignorent que leur langue sera analysée, condition sine qua non pour garantir l’authenticité des données. L’augmentation des textes générés par l’IA remet en question cette conception traditionnelle des corpus comme archives de l’usage authentique de la langue.
Alors que les frontières entre la langue produite par l’homme et celle générée par la machine deviennent de plus en plus floues, plusieurs questions se posent : quel statut donner aux textes générés par l’IA ? Comment les distinguer des productions humaines ? Quelles implications pour notre compréhension du langage et son évolution ? Comment endiguer la contamination potentielle des données destinées à l’étude linguistique ?
Une langue moyenne et désincarnée
On peut parfois avoir l’illusion de converser avec un humain, comme dans le film « Her » (2013), mais c’est une illusion. L’IA, alimentée par nos instructions (les fameux « prompts »), manipule des millions de données pour générer des suites de mots probables, sans réelle compréhension humaine. Notre IA actuelle n’a pas la richesse d’une voix humaine. Son style est reconnaissable parce que moyen. C’est le style de beaucoup de monde, donc de personne.
Bande annonce du film « Her » (2013) de Spike Jonze.
À partir d’expressions issues d’innombrables textes, l’IA calcule une langue moyenne. Le processus commence par un vaste corpus de données textuelles qui rassemble un large éventail de styles linguistiques, de sujets et de contextes. Au fur et à mesure l’IA s’entraîne et affine sa « compréhension » de la langue (par compréhension, il faut entendre la connaissance du voisinage des mots) mais en atténue ce qui rend chaque manière de parler unique. L’IA prédit les mots les plus courants et perd ainsi l’originalité de chaque voix.
Bien que ChatGPT puisse imiter des accents et des dialectes (avec un risque de caricature), et changer de style sur demande, quel est l’intérêt d’étudier une imitation sans lien fiable avec des expériences humaines authentiques ? Quel sens y a-t-il à généraliser à partir d’une langue artificielle, fruit d’une généralisation déshumanisée ?
Parce que la linguistique relève des sciences humaines et que les phénomènes grammaticaux que nous étudions sont intrinsèquement humains, notre mission de linguistes exige d’étudier des textes authentiquement humains, connectés à des expériences humaines et des contextes sociaux. Contrairement aux sciences exactes, nous valorisons autant les régularités que les irrégularités langagières. Prenons l’exemple révélateur de l’expression « après que » : normalement suivie de l’indicatif, selon les livres de grammaire, mais fréquemment employée avec le subjonctif dans l’usage courant. Ces écarts à la norme illustrent parfaitement la nature sociale et humaine du langage.
La menace de l’ouroboros
La contamination des ensembles de données linguistiques par du contenu généré par l’IA pose de grands défis méthodologiques. Le danger le plus insidieux dans ce scénario est l’émergence de ce que l’on pourrait appeler un « ouroboros linguistique » : un cycle d’auto-consommation dans lequel les grands modèles de langage apprennent à partir de textes qu’ils ont eux-mêmes produits.
Cette boucle d’autorenforcement pourrait conduire à une distorsion progressive de ce que nous considérons comme le langage naturel, puisque chaque génération de modèles d’IA apprend des artefacts et des biais de ses prédécesseurs et les amplifie.
Il pourrait en résulter un éloignement progressif des modèles de langage humain authentique, ce qui créerait une sorte de « vallée de l’étrange » linguistique où le texte généré par l’IA deviendrait simultanément plus répandu et moins représentatif d’une communication humaine authentique."
https://theconversation.com/vocabulaire-et-diversite-linguistique-comment-lia-appauvrit-le-langage-252944
#metaglossia_mundus
"...BookTranslate.ai is the Gutenberg Revolution of translation.”— Balint Taborski EGER, HUNGARY, June 5, 2025 /EINPresswire.com/ -- BookTranslate.ai, a groundbreaking AI-powered literary translation platform, has outperformed legendary human translator Donald Keene in a blind test — and it’s now open to publishers, authors, and translators worldwide.
Unlike generic AI tools, BookTranslate.ai doesn’t just translate — it analyzes, refines, and critiques its own output across three advanced stages. The system includes a sophisticated AI Literary Analysis Engine, an Advanced Multi-Pass Refinement System, and a final-stage AI Review Engine (“The Finalizer”). This full-stack process delivers translations of exceptional fidelity and style — even surpassing acclaimed human translations in blind literary evaluations.
Founded by Hungarian indie publisher, translator, and software developer Balint Taborski (Praxeum Publishing), BookTranslate.ai emerged from the critical need for high-fidelity translations of complex literary and philosophical works. "Traditional translation is often prohibitively expensive and slow for small, niche, budget-constrained publishers like myself, while generic machine translation tools lacked the nuance for publishable quality," says Taborski. "I needed to cut costs without sacrificing quality. I needed a system that could not only translate with precision but also preserve authorial intent and literary integrity. What began as an internal tool has evolved into a platform that can empower authors and publishers worldwide."
BookTranslate.ai's unique three-stage process ensures unparalleled quality:
1. AI Literary Analysis Engine (Preprocessor): Before translation, this engine performs a deep, holistic analysis of the entire manuscript, understanding its genre, style, themes, and structure. It generates a custom "Translation Blueprint" – detailed instructions and an AI-suggested glossary – to guide the subsequent translation with unparalleled contextual awareness.
2. Advanced Multi-Pass Refinement System: Each paragraph, guided by the Blueprint, undergoes five distinct AI-driven passes of translation and proofreading. This iterative process refines grammar, fluency, tone, and stylistic consistency, aiming for output that is ~98% publication-ready.
3. AI Final Review Engine (The "Finalizer"): This final stage acts as an expert AI second opinion. The Finalizer meticulously compares the AI-translated text against the original source, identifying subtle errors, inconsistencies, or nuanced improvement opportunities that even a robust multi-pass system might overlook. Its suggestions, which users can review and selectively apply, further elevate the translation's polish towards perfection.
The platform's capabilities have been validated through multiple public demonstrations and blind tests. Notably, in an evaluation of Osamu Dazai's Ningen Shikkaku (No Longer Human), Google's Gemini 2.5 Pro model assessed BookTranslate.ai's translation as "Best," outperforming both DeepL and Donald Keene's acclaimed version, stating it "delivered the most fluent, faithful, and emotionally rich translation." Gemini also remarked, when reviewing BookTranslate.ai's unedited English translation of a 19th century classic of French political philosophy, Édouard Laboulaye’s *The State and Its Limits*, "Given the consistency and elegance, it's more straightforward to attribute it to a talented human translator."
Additional compelling case studies, including blind analyses of translations for works by H.P. Lovecraft and Eugen von Böhm-Bawerk, demonstrating the system's versatility and the AI Finalizer's impact, are available on the BookTranslate.ai website.
"BookTranslate.ai is the Gutenberg Revolution of translation," Taborski adds. "It's not just about AI generating text; it's about AI demonstrating a deep literary understanding, refining its own work with meticulous care, and even providing critical editorial feedback. An entire AI editorial board is collaborating on the manuscript under the hood. This allows us to make great literature and important ideas accessible across languages at a fraction of the cost, without sacrificing quality. Soon, we will look upon traditional human translation much as we now look upon the hand-copying of books."
BookTranslate.ai invites authors, publishers, and translators to experience the future of literary translation. Full book examples, a free trial for up to 3000 characters, and an on-request full-chapter translation offer are available at https://booktranslate.ai.
About BookTranslate.ai:
BookTranslate.ai is an advanced AI translation platform specializing in books and long-form texts. Its proprietary three-stage system—comprising pre-translation literary analysis, multi-pass AI refinement, and an AI final review engine—delivers publishable-quality translations that preserve authorial voice and nuance. Founded by indie publisher Balint Taborski, BookTranslate.ai aims to democratize access to global literature and ideas.
Balint Taborski
BookTranslate.ai
+36 70 218 1428
hello@booktranslate.ai
https://www.pahomepage.com/business/press-releases/ein-presswire/819508395/booktranslate-ai-stuns-publishing-world-ai-outperforms-legendary-translator-in-blind-test/
#metaglossia_mundus
"TransPerfect Launches New App for Mobile Interpretation
Supports Intuitive Access to a Phone or Video Interpreter in Seconds
NEW YORK, June 05, 2025 (GLOBE NEWSWIRE) -- TransPerfect, the world's largest provider of language and AI solutions for global business, today announced it has launched the TransPerfect Interpretation App, enabling users to connect to a live, expertly trained video or phone interpreter in seconds.
Available on the web, iOS, Android, and Microsoft devices, the app makes accessing over-the-phone interpretation and video remote interpretation simple and fast. It supports 200+ languages, including American Sign Language, enabling users across multiple industries to quickly and effectively respond to customer needs. It’s ideal for hospitals, medical and benefit providers, financial branches, legal, retail stores, first responders, and others in any situation where customer interactions happen outside of a contact center.
“This app marks a significant leap forward in making interpretation services more accessible and easier to use,” commented TransPerfect Connect Vice President Steven Cheeseman. “With on-demand access to professional interpreters, we’re empowering organizations to overcome language barriers in real time. It’s fast, intuitive, and purpose-built to support the critical work our clients perform every day.”
To use the app, simply select a language from the menu and tap to choose audio or video interpretation. If a video interpreter is not available, the call will automatically route to an audio interpreter.
Available to all existing TransPerfect customers, the app enables users to:
Capture necessary data (account numbers, claim numbers, medical record numbers, etc.)
Adjust the audio, mute lines, turn video on and off
Add additional parties regardless of whether they have the application installed
Capture user satisfaction at the end of every call (customizable by client)
TransPerfect President and Co-CEO Phil Shawe stated, “TransPerfect customers now have easy access to a live interpreter—wherever, whenever, and in any language.”
To learn more about the launch, visit https://go.transperfect.com/mobile_interpretation or email us at mobileinterp@transperfect.com..."
June 05, 2025 10:00 ET
| Source: TransPerfect
https://www.globenewswire.com/news-release/2025/06/05/3094498/0/en/TransPerfect-Launches-New-App-for-Mobile-Interpretation.html
#metaglossia_mundus
Google reveals details behind the development of AI Mode. Learn design insights and what marketers need to know.
Google built AI Mode to handle longer, conversational search queries.
Google claims AI search delivers "more qualified clicks."
Google has no data to share to verify claims of quality improvements.
SEJ STAFF
Matt G. Southern
Google has shared new details about how it designed and built AI Mode.
In a blog post, the company reveals the user research, design challenges, and testing that shaped its advanced AI search experience.
These insights may help you understand how Google creates AI-powered search tools. The details show Google’s shift from traditional keyword searches to natural language conversations.
User Behavior Drove AI Mode Creation
Google built AI Mode in response to the ways people were using AI Overviews.
Google’s research showed a disconnect between what searchers wanted and what was available.
Claudia Smith, UX Research Director at Google, explains:
“People saw the value in AI Overviews, but they didn’t know when they’d appear. They wanted them to be more predictable.”
The research also found people started asking longer questions. Traditional search wasn’t built to handle these types of queries well.
This shift in search behavior led to a question that drove AI Mode’s creation, explains Product Management Director Soufi Esmaeilzadeh:
“How do you reimagine a Search gen AI experience? What would that look like?”
AI “Power Users” Guided Development Process
Google’s UX research team identified the most important use cases as: exploratory advice, how-to guides, and local shopping assistance.
This insight helped the team understand what people wanted from AI-powered search.
Esmaeilzadeh explained the difference:
“Instead of relying on keywords, you can now pose complex questions in plain language, mirroring how you’d naturally express yourself.”
According to Esmaeilzadeh, early feedback suggests that the team’s approach was successful:
“They appreciate us not just finding information, but actively helping them organize and understand it in a highly consumable way, with help from our most intelligent AI models.”
Industry Concerns Around AI Mode
While Google presents an optimistic development story, industry experts are raising valid concerns.
John Shehata, founder of NewzDash, reports that sites are already “losing anywhere from 25 to 32% of all their traffic because of the new AI Overviews.” For news publishers, health queries show 26% AI Overview penetration.
Mordy Oberstein, founder of Unify Brand Marketing, analyzed Google’s I/O demonstration and found the examples weren’t as complex as presented. He shows how Google combined readily available information rather than showcasing advanced AI reasoning.
Google’s claims about improved user engagement have not been verified. During a recent press session, Google executives claimed AI search delivers “more qualified clicks” but admitted they have “no data to share” on these quality improvements.
Further, Google’s reporting systems don’t differentiate between clicks from traditional search, AI overviews, and AI mode. This makes independent verification impossible.
Shehata believes that the fundamental relationship between search and publishers is changing:
“The original model was Google: ‘Hey, we will show one or two lines from your article, and then we will give you back the traffic. You can monetize it over there.’ This agreement is broken now.”
What This Means
For SEO professionals and content marketers, Google’s insights reveal important changes ahead.
The shift from keyword targeting to conversational queries means content strategies need to focus on directly answering user questions rather than optimizing for specific terms.
The focus on exploratory advice, how-to content, and local help shows these content types may become more important in AI Mode results.
Shehata recommends that publishers focus on content with “deep analysis of a situation or an event” rather than commodity news that’s “available on hundreds and thousands of sites.”
He also notes a shift in success metrics: “Visibility, not traffic, is the new metric” because “in the new world, we will get less traffic.”
Looking Ahead
Esmaeilzadeh said significant work continues:
“We’re proud of the progress we’ve made, but we know there’s still a lot of work to do, and this user-centric approach will help us get there.”
Google confirmed that more AI Mode features shown at I/O 2025 will roll out in the coming weeks and months. This suggests the interface will keep evolving based on user feedback and usage patterns."
Matt G. Southern
Senior News Writer at Search Engine Journal
https://www.searchenginejournal.com/google-shares-details-behind-ai-mode-development/548474/
#metaglossia_mundus
Ngũgĩ’s most important contributions were on language. His novels, his essays, and his unyielding belief in the dignity of African languages reshaped world literature. To call him merely a “great African writer” would be to shrink his genius – he was one of the most vital thinkers of our age, a voice who spoke from Kenya but to humanity.
"Ngũgĩ wa Thiong’o: The Writer Who Made Language a Battlefield
Ngũgĩ’s most important contributions were on language. His novels, his essays, and his unyielding belief in the dignity of African languages reshaped world literature. To call him merely a “great African writer” would be to shrink his genius – he was one of the most vital thinkers of our age, a voice who spoke from Kenya but to humanity.
Ngũgĩ wa Thiong’o, the literary giant from Kenya and unflinching advocate for African languages, passed away on 28 May 2025, leaving behind a legacy that bridged continents and generations. For those of us who knew him – whether through his words or in solidarity campaigns – his life was a testament to the power of art as resistance.
I am deeply saddened to learn of Ngũgĩ’s passing. Ngũgĩ did not just write books – he declared war. War on colonialism and neocolonialism, on linguistic shame, on the very idea that African thought must be filtered through colonial grammar to be “literature”. Always assuming that we would soon have our next conversation, I felt bereft when news of his death broke.
When I helped organise London protests for his release in 1978, we shouted his banned titles outside Kenya’s High Commission. We prepared banners saying “Kenyatta sheds Petals of Blood”, Waingereza walifunga Kimathi, Mzee anafunga Ngũgĩ! (The British locked up Kimathi, Mzee [Kenyatta] locks up Ngũgĩ!). Years later, when I visited him at his home near Kamĩrĩĩthu, the site of the people’s theatre at which Ngaahika Ndeenda was performed, and which the regime destroyed, he told me: “You thought you were demanding my freedom? You were demanding yours.” He showed me the manuscript he’d written on scraps of paper while in prison. “They gave me a pen to confess,” he laughed. “I used it to imagine”. “They thought they buried me in prison,” he laughed, “but they planted a seed”. His warmth and clarity in person mirrored the fearlessness of his writing.
I always resented the fact that so much of his writing was relegated to the African Writers Series. True, the series brought attention to the wealth of writers from the African continent. But many of them were, like Ngũgĩ, literary giants, not merely “African writers”. Their contribution to art, politics, and thinking were and are of universal relevance. The failure to recognise this universalism is perhaps one of the reasons that, despite being nominated several times, Ngũgĩ was denied the Nobel prize for literature.
Ngũgĩ’s most important contributions were on language. His novels, his essays, and his unyielding belief in the dignity of African languages reshaped world literature. To call him merely a “great African writer” would be to shrink his genius – he was one of the most vital thinkers of our age, a voice who spoke from Kenya but to humanity. Ngũgĩ’s decision to write in Gĩkũyũ was not symbolic – it was an act of intellectual insurrection. He tore down the lie that creativity required the approval of imperial languages. “To think in my mother tongue,” he once told me, “is to dream in freedom”. His Gĩkũyũ works (Mũrogi wa Kagogo, Matigari) were not “translations from English” – they were the originals, the canon itself. Beyond “African Writer”, Western obituaries will box him into “African literature”. But Ngũgĩ belonged to the same pantheon as Dostoevsky, Marquez, and Orwell – writers who exposed the machinery of power. Wizard of the Crow is as universal as 1984; Decolonising the Mind, perhaps as urgent as Fanon.
In his Foreword to the Daraja Press edition of Mau Mau From Within: The Story of the Kenya Land and Freedom Army, Ngũgĩ wrote: “We don’t have to use the vocabulary of the colonial to describe our struggles, especially today when there is a worldwide movement to overturn monuments to slavery and colonialism.”
For those of us who stood beside Ngũgĩ – in protest, in lecture halls, or in his Limuru home – he was also a generous comrade, always leaning forward to listen, even as the world often leaned away from him. I admired how attentive Ngũgĩ always was to aspiring younger writers. At conferences where we both spoke, I watched him corner young African writers, urging them to “write dangerously”. He’d ask: “Who owns your language? Who profits from your silence?” His belief in them was visceral. “Talent is common,” he told them, “What’s rare is the courage to use it.”
In his piercing 2020 interview with Daraja Press, Ngũgĩ laid bare his life’s mission: Colonialism” stole our land, but language was the theft of our dreams”. He recounted how, as a child, British teachers beat Gĩkũyũ out of his classmates–”not just with canes, but with the lie that our words were small, ugly things”. His entire oeuvre, from his early writings as “James” Ngugi, to Petals of Blood, Decolonising the Mind and Wizard of the Crow, was a counterattack: “I write in Gĩkũyũ not to be provincial, but to prove that the universal lives in the particular.”
Kenya today claims Ngũgĩ as a national treasure – yet how recklessly it squandered him in his prime. No streets are named for the man whose books were banned in schools, politicians quote Decolonising the Mind while gutting funding for indigenous-language education. When I last visited Nairobi, a young poet told me, “They praise Ngũgĩ now because he’s no longer dangerous.” But Ngũgĩ’s legacy is danger: the kind that outlives regimes.
Kiumbi kiaguo ikio ki, kiaguo ikio. “A thing that forgets its origins will soon die.” Ngũgĩ, they jailed you, exiled you, misnamed you – yet here you remain.
Rest well, comrade and friend.
Sincere condolences to Ngugi’s family."
Firoze Manji
June 5, 2025
https://www.theelephant.info/opinion/2025/06/05/ngugi-wa-thiongo-the-writer-who-made-language-a-battlefield/
#metaglossia_mundus
En rompant avec l’anglais pour écrire en kikuyu, il a incarné une pensée radicale de la décolonisation.
"L'héritage vivant de Ngũgĩ wa Thiong’o : décoloniser l’esprit et les langues
Published: June 5, 2025 3.59pm SAST
Christophe Premat, Stockholm University
Ngũgĩ wa Thiong’o est mort le 28 mai 2025 à l’âge de 87 ans. Son nom restera dans l’histoire non seulement comme celui d’un grand romancier kenyan, mais aussi comme celui d’un penseur radical de la décolonisation. À l’instar de Valentin-Yves Mudimbe, disparu quelques semaines plus tôt, il a su interroger les conditions mêmes de la possibilité d’un savoir africain en contexte postcolonial. Mais là où Mudimbe scrutait les « bibliothèques coloniales » pour en dévoiler les présupposés, Ngũgĩ a voulu transformer la pratique même de l’écriture : en cessant d’écrire en anglais pour privilégier sa langue maternelle, le kikuyu, il a posé un geste politique fort, un acte de rupture.
En tant que spécialiste des théories postcoloniales, j'analyse la manière dont ces parcours critiques se sont efforcés à repenser la manière dont le savoir est produit et transmis en Afrique.
Pour Ngũgĩ, la domination coloniale ne s’arrête pas aux frontières, aux institutions ou aux lois. Elle prend racine dans les structures mentales, dans la manière dont un peuple se représente lui-même, ses valeurs, son passé, son avenir.
Langue et pouvoir : une géopolitique de l’imaginaire
Dans Décoloniser l’esprit (1986), Ngũgĩ wa Thiong’o explique pourquoi il a décidé d’abandonner l’anglais, langue dans laquelle il avait pourtant connu un succès international. Il y pose une affirmation devenue centrale dans les débats sur les héritages coloniaux : « Les vrais puissants sont ceux qui savent leur langue maternelle et apprennent à parler, en même temps, la langue du pouvoir. » Car tant que les Africains seront contraints de penser, de rêver, d’écrire dans une langue qui leur a été imposée, la libération restera incomplète.
À travers la langue, les colonisateurs ont conquis bien plus que des terres : ils ont imposé une certaine vision du monde. En contrôlant les mots, ils ont contrôlé les symboles, les récits, les hiérarchies culturelles. Pour Ngũgĩ, le bilinguisme colonial ne relève pas d’un enrichissement, mais d’une fracture : il sépare la langue du quotidien (la langue vernaculaire) de celle de l’école, de la pensée, du droit, de la littérature. Il y voit une violence structurelle, une « dissociation entre l’esprit et le corps », qui rend impossible une appropriation pleine et entière de l’expérience africaine.
Une aliénation tenace
L’analyse de Ngũgĩ met en lumière les impasses des politiques linguistiques postcoloniales qui ont souvent continué à faire des langues européennes des langues d’État, de savoir et de prestige, tout en reléguant les langues africaines à la sphère privée. C’est en ce sens qu’on peut parler de diglossie, c’est-à-dire de situation de cohabitation de deux langues avec des fonctions sociales distinctes. Cette diglossie instituée produit une hiérarchisation des langues qui reflète, en profondeur, une hiérarchisation des cultures.
We’re 10! Support us to keep trusted journalism free for all.
Support our cause
Loin d’en appeler à un retour passéiste ou à une clôture identitaire, Ngũgĩ veut libérer le potentiel des langues africaines : leur permettre de dire le contemporain, d’inventer une modernité qui ne soit pas un simple calque des modèles européens. Il reprend ici à son compte la tâche historique que se sont donnée les écrivains dans d’autres langues « mineures » : faire pour le kikuyu ce que Shakespeare a fait pour l’anglais, ou Tolstoï pour le russe.
Il s’agit non seulement d’écrire dans les langues africaines, mais de faire en sorte que ces langues deviennent des vecteurs de philosophie, de sciences, d’institutions — bref, de civilisation. Ce choix d’écrire en kikuyu ne fut pas sans conséquences. En 1977, Ngũgĩ coécrit avec Ngũgĩ wa Mirii une pièce de théâtre, Ngaahika Ndeenda (« Je me marierai quand je voudrai »), jouée en langue kikuyu dans un théâtre communautaire de Kamiriithu, près de Nairobi.
La pièce, portée par des acteurs non professionnels, dénonçait avec virulence les inégalités sociales et les survivances du colonialisme au Kenya. Son succès populaire inquiète les autorités : quelques semaines après la première, Ngũgĩ est arrêté sans procès et incarcéré pendant près d’un an. À sa libération, interdit d’enseigner et surveillé de près, il choisit l’exil.
Ce bannissement de fait durera plus de vingt ans. C’est dans cette période de rupture qu’il entame l’écriture en kikuyu de son roman Caitaani Mutharaba-Ini (Le Diable sur la croix), qu’il rédige en prison sur du papier hygiénique.
Une pensée toujours actuelle
L’œuvre de Ngũgĩ éclaire la manière dont les sociétés africaines contemporaines restent prises dans des logiques de domination symbolique, malgré les indépendances politiques. La mondialisation a remplacé les formes les plus brutales de l’impérialisme, mais elle reconduit souvent les logiques de domination symbolique. Dans le champ culturel, les ex-puissances coloniales continuent d’exercer une influence considérable à travers les réseaux diplomatiques, éducatifs, éditoriaux.
La Francophonie, par exemple, se présente comme un espace de coopération linguistique, mais elle perpétue souvent des asymétries dans la validation des productions culturelles. Le fait de présenter les langues coloniales comme des langues de communication dépassant les clôtures des langues vernaculaires est une illusion que Ngũgĩ dénonce avec virulence.
Des penseurs comme Jean-Godefroy Bidima ou Seloua Luste Boulbina ont montré à quel point les politiques linguistiques postcoloniales ont tendance à officialiser certaines langues au détriment d’autres, créant une nouvelle forme de langue de bois, souvent coupée des réalités populaires. La philosophe algéro-française évoque à ce propos un espace public plurilingue à instituer: un espace qui ne se résume pas à opposer langues coloniales et langues vernaculaires, mais qui réinvente les usages, les hybridations, les ruses du langage.
Cette réflexion fait écho à la position de Ngũgĩ : écrire dans sa langue ne suffit pas, encore faut-il produire un langage à la hauteur des luttes sociales et politiques.
Pour une mémoire active de Ngũgĩ wa Thiong’o
À l’heure où les débats sur la décolonisation se multiplient, souvent vidés de leur substance ou récupérés par des logiques institutionnelles, relire Ngũgĩ wa Thiong’o permet de revenir à l’essentiel : l’émancipation passe par un changement de regard sur soi-même, qui commence dans la langue. La véritable libération n’est pas seulement politique ou économique ; elle est aussi culturelle, cognitive, symbolique.
En refusant de penser depuis des catégories importées, en assumant le risque d’un geste radical, Ngũgĩ wa Thiong’o a ouvert la voie à une pensée authentiquement africaine, enracinée et universelle. Son œuvre rappelle qu’il ne suffit pas de parler au nom de l’Afrique ; encore faut-il parler depuis elle, avec ses langues, ses imaginaires, ses luttes. À l’heure de sa disparition, son message reste plus vivant que jamais."
https://theconversation.com/lheritage-vivant-de-ngugi-wa-thiongo-decoloniser-lesprit-et-les-langues-258209
#metaglossia_mundus
"New York, the World’s Most Linguistically Diverse Metropolis
In Language City, Ross Perlin, a linguist, takes readers on a tour of the city’s communities with endangered tongues.
June 05, 2025
As co-director of the Endangered Language Alliance, Ross Perlin has managed a variety of projects focused on language documentation, language policy, and public programming around urban linguistic diversity. Photo by Cecil Howell.
Half of all 7,000-plus human languages may disappear over the next century, and—because many have never been recorded—when they’re gone, it will be forever. Ross Perlin, a lecturer in Columbia’s Department of Slavic Languages and co-director of the Endangered Language Alliance, is racing against time to map little-known languages across New York. In his new book, Language City, Perlin follows six speakers of endangered languages deep into their communities, from the streets of Brooklyn and Queens to villages on the other side of the world, to learn how they are maintaining and reviving their languages against overwhelming odds. He explores the languages themselves, from rare sounds to sentence-long words to bits of grammar that encode entirely different worldviews.
Seke is spoken by 700 people from five ancestral villages in Nepal, and a hundred others living in a single Brooklyn apartment building. N’ko is a radical new West African writing system now going global in Harlem and the Bronx. After centuries of colonization and displacement, Lenape, the city’s original indigenous language and the source of the name Manhattan, “the place where we get bows,” has just one native speaker, along with a small band of revivalists. Also profiled in the book are speakers of the indigenous Mexican language Nahuatl, the Central Asian minority language Wakhi, and Yiddish, braided alongside Perlin’s own complicated family legacy. On the 100th anniversary of a notorious anti-immigration law that closed America’s doors for decades, and the 400th anniversary of New York’s colonial founding, Perlin raises the alarm about growing political threats and the onslaught of languages like English and Spanish.
Perlin talks about the book with Columbia News, along with why New York is so linguistically diverse, and his current and upcoming projects.
How did this book come about?
I’m a linguist, writer, and translator, and have been focused on language endangerment and documentation for the last two decades. Initially, I worked in southwest China on a dictionary, a descriptive grammar, and corpus of texts for Trung, a language spoken in a single remote valley on the Burmese border. While finishing up my dissertation, I came back to New York and joined the Endangered Language Alliance (ELA), which had recently been founded to focus on urban linguistic diversity.
ELA’s key insight was a paradox: Even as language loss accelerates everywhere—with as many as half of the world’s 7,000 languages now considered endangered—cities are more linguistically diverse than ever before, thanks to migration, urbanization, and diaspora. This means that speakers of endangered, indigenous, and primarily oral languages are now often right next door, and that fieldwork can happen in a different key, with linguists and communities making common cause as neighbors, for the long term.
For the last 12 years, I’ve been ELA’s co-director, together with Daniel Kaufman. Leading a small, independent nonprofit has been its own adventure. We do everything from in-depth research on little-studied languages (projects on Jewish, Himalayan, and indigenous Latin American languages, for example, with much of the research happening right here in the five boroughs) to public events, classes, collaborations with city agencies, childrens books, and the first-ever language map of New York City. At some point, I knew that all my notes on everything I was learning and seeing around the city and beyond had to become a book. The urgency only grew as a series of unprecedented crises started hitting the immigrant and diaspora communities we work with. The crises are still unfolding.
Can you share some details about the six people you portray in the book, and the endangered languages they speak?
Language City is both the story of New York’s languages—the past, present, and future of the world's most linguistically diverse city—and the story of six specific people doing extraordinary things to keep their embattled mother tongues alive. They come from all over the world, but converge in New York. They represent a variety of strategies for language maintenance and revitalization in the face of tremendous odds. All of them are people I’ve known and worked with for years. Between them, in all their multilingualism, they actually speak about 30 languages, but they are also regular people you might run into on the subway.
Rasmina is one of the youngest speakers of Seke, a language from five villages in Nepal, and now a sixth, vertical village in the middle of Brooklyn. Husniya, a speaker of Wakhi from Tajikistan, has gone through every stage of her life and education in a different language, and can move easily along New York City’s new Silk Road. Originally from Moldova, Boris is a Yiddish novelist, poet, editor, and one-man linguistic infrastructure.
Ibrahima, a language activist from Guinea, champions N’ko, a relatively new writing system created to challenge the dominance of colonial languages in West Africa. Irwin is a Queens chef who writes poetry in, and cooks through, his native Nahuatl, the language of the Aztecs. And Karen is a keeper of Lenape, the original language of the land on which New York was settled.
How did New York end up as such a linguistically diverse city?
Four hundred years ago, this Lenape-speaking archipelago became nominally Dutch under the West India Company. In fact, the territory evolved into an unusual commercial entrepôt at the fulcrum of three continents, where the languages of Native Americans, enslaved Africans, and European refugees and traders were all in the mix. A reported 18 languages were spoken by the first 400-500 inhabitants. This set the template, but the defining immigration waves of the 19th and 20th centuries were of another order, as New York became a global center for business, politics, and culture, as well as the pre-eminent gateway to the U.S. and a bridge between hemispheres. From the half-remembered myth of Ellis Island to the very present reality of 200,000 asylum seekers arriving in the last few years, I argue—in what I think is the first linguistic history of any city—that applying the lens of language helps us understand the city (and all cities) in profoundly new ways. From Frisian and Flemish 400 years ago, to Fujianese and Fulani today, deep linguistic diversity, although always overlooked, has been fundamental.
What did you teach in the spring semester?
The class I’ve taught every year for the last six years—Endangered Languages in the Global City. I designed it around our research at ELA, and it’s completely unique to Columbia. Hundreds of students sign up, testifying to the massive interest Columbia students have in learning about linguistic diversity, though we can admit only a fraction of them. Many of the endangered language speakers and activists featured in Language City have visited the class, and we also take students out to some of the city’s most linguistically diverse neighborhoods, where many students have done strikingly original fieldwork.
The spring semester was something of a departure: I stepped in to teach Language and Society, as well as a course I designed last year called Languages of Asia, which focuses on the continent’s lesser-known language families and linguistic areas. I’ve also been supervising senior theses together with Meredith Landman,who directs the undergraduate program in linguistics. Across the first half of the 20th century, Columbia and Barnard became foundational sites for the study of linguistic diversity and for documenting languages, thanks to Franz Boas, who, in 1902, became the head of Columbia’s anthropology department—the first in the country—and his students. There is hope today for Columbia’s formative role to be revived: The University's local and global position makes this as urgent and achievable as ever. Additionally, the linguistics major was recently (at last) restored, and an ever-growing number of students are doing remarkable work on languages from around the world.
What are you working on now?
ELA keeps on ELA’ing, with a huge range of projects. On any given day, we might be recording and working with speakers of languages originally from Guatemala, Nepal, or Iran. With 700+ languages and counting, our language map (languagemap.nyc) is continually being updated, and we have people here making connections between language and health, education, literacy, technology, translation. This fall, I’ll be in Berlin as a fellow of the American Academy, researching via an obscure film (in 2,000 languages and counting) how missionary linguists and Bible translators shape global linguistic diversity.
Any summer plans related to language preservation at Columbia or elsewhere?
Ongoing work at ELA: The sky, if not for the budget, would be the limit. There have been recent sessions on a little-known language from Afghanistan, with a woman who is probably the only speaker in New York. Someone whose family still remembers a highly endangered Jewish language variant from Iran recently got in touch. And we met a speaker of a Mongolic language of western China at a restaurant. I also hope to continue soon some scouting of urban linguistic diversity in the Caucasus, past and present, with support from the Harriman Institute."
https://news.columbia.edu/news/new-york-worlds-most-linguistically-diverse-metropolis
#metaglossia_mundus
"Newswise — Creativity often emerges from the interplay of disparate ideas—a phenomenon known as combinational creativity. Traditionally, tools like brainstorming, mind mapping, and analogical thinking have guided this process. Generative Artificial Intelligence (AI) introduces new avenues: large language models (LLMs) offer abstract conceptual blending, while image (T2I) and T2-three-dimensional (3D) models turn text prompts into vivid visuals or spatial forms. Yet despite their growing use, little research has clarified how these tools function across different stages of creativity. Without a clear framework, designers are left guessing which AI tool fits best. Given this uncertainty, in-depth studies are needed to evaluate how various AI dimensions contribute to the creative process.
A research team from Imperial College London, the University of Exeter, and Zhejiang University has tackled this gap. Their new study (DOI: 10.1016/j.daai.2025.100006), published in May 2025 in Design and Artificial Intelligence, investigates how generative AI models with different dimensional outputs support combinational creativity. Through two empirical studies involving expert and student designers, the team compared the performance of LLMs, T2I, and T2-3D models across ideation, visualization, and prototyping tasks. The results provide a practical framework for optimizing human-AI collaboration in real-world creative settings.
To map AI's creative potential, the researchers first asked expert designers to apply each AI type to six combinational tasks—including splicing, fusion, and deformation. LLMs performed best in linguistic-based combinations such as interpolation and replacement but struggled with spatial tasks. In contrast, T2I and T2-3D excelled at visual manipulations, with 3D models especially adept at physical deformation. In a second study, 24 design students used one AI type to complete a chair design challenge. Those using LLMs generated more conceptual ideas during early, divergent phases but lacked visual clarity. T2I models helped externalize these ideas into sketches, while T2-3D tools offered robust support for building and evaluating physical prototypes. The results suggest that each AI type offers unique strengths, and the key lies in aligning the right tool with the right phase of the creative process.
“Understanding how different generative AI models influence creativity allows us to be more intentional in their application,” said Prof. Peter Childs, co-author and design engineering expert at Imperial College London. “Our findings suggest that large language models are better suited to stimulate early-stage ideation, while text-to-image and text-to-3D tools are ideal for visualizing and validating ideas. This study helps developers and designers align AI capabilities with the creative process rather than using them as one-size-fits-all solutions.”
The study's insights are poised to reshape creative workflows across industries. Designers can now match AI tools to specific phases—LLMs for generating diverse concepts, T2I for rapidly visualizing designs, and T2-3D for translating ideas into functional prototypes. For educators and AI developers, the findings provide a blueprint for building more effective, phase-specific design tools. By focusing on each model’s unique problem-solving capabilities, this research elevates the conversation around human–AI collaboration and paves the way for smarter, more adaptive creative ecosystems.
###
References
DOI
10.1016/j.daai.2025.100006
Original Source URL
https://doi.org/10.1016/j.daai.2025.100006
Funding information
The first and third authors would like to acknowledge the China Scholarship Council (CSC)."
Released: 5-Jun-2025 7:35 AM EDT
Source Newsroom: Chinese Academy of Sciences
https://www.newswise.com/articles/designing-with-dimensions-rethinking-creativity-through-generative-ai
##metaglossia_mundus
"AI reveals hidden language patterns and likely authorship in the Bible
by Duke University
edited by Sadie Harley, reviewed by Robert Egan
AI is transforming every industry, from medicine to film to finance. So, why not use it to study one of the world's most revered ancient texts, the Bible?
An international team of researchers, including Shira Faigenbaum-Golovin, assistant research professor of Mathematics at Duke University, combined artificial intelligence, statistical modeling and linguistic analysis to address one of the most enduring questions in biblical studies: the identification of its authors.
The study is published in the journal PLOS One.
By analyzing subtle variations in word usage across texts, the team was able to distinguish between three distinct scribal traditions (writing styles) spanning the first nine books of the Hebrew Bible, known as the Enneateuch.
Using the same AI-based statistical model, the team was then able to determine the most likely authorship of other Bible chapters. Even better, the model also explained how it reached its conclusions.
But how did the mathematician get here?
In 2010, Faigenbaum-Golovin began collaborating with Israel Finkelstein, head of the School of Archaeology and Maritime Cultures at the University of Haifa, using mathematical and statistical tools to determine the authorship of lettering found on pottery fragments from 600 B.C. by comparing the style and shape of the letters inscribed on each fragment.
Their discoveries were featured on the front page of The New York Times.
"We concluded that the findings in those inscriptions could offer valuable clues for dating texts from the Old Testament," Faigenbaum-Golovin said. "That's when we started putting together our current team, who could help us analyze these biblical texts."
The multidisciplinary undertaking was made up of two parts. First, Faigenbaum-Golovin and Finkelstein's team—Alon Kipnis (Reichman University), Axel Bühler (Protestant Faculty of Theology of Paris), Eli Piasetzky (Tel Aviv University) and Thomas Römer (Collège de France)—was made up of archaeologists, biblical scholars, physicists, mathematicians and computer scientists.
The team used a novel AI-based statistical model to analyze language patterns in three major sections of the Bible. They studied the Bible's first five books: Deuteronomy, the so-called Deuteronomistic History from Joshua to Kings, and the priestly writings in the Torah.
Results showed Deuteronomy and the historical books were more similar to each other than to the priestly texts, which is already the consensus among biblical scholars.
"We found that each group of authors has a different style—surprisingly, even regarding simple and common words such as 'no,' 'which,' or 'king.' Our method accurately identifies these differences," said Römer.
To test the model, the team selected 50 chapters from the first nine books of the Bible, each of which has already been allocated by biblical scholars to one of the writing styles mentioned above.
"The model compared the chapters and proposed a quantitative formula for allocating each chapter to one of the three writing styles," said Faigenbaum-Golovin.
In the second part of the study, the team applied their model to chapters of the Bible whose authorship was more hotly debated. By comparing these chapters to each of the three writing styles, the model was able to determine which group of authors was more likely to have written them. Even better: the model also explained why it was making these calls.
"One of the main advantages of the method is its ability to explain the results of the analysis—that is, to specify the words or phrases that led to the allocation of a given chapter to a particular writing style," said Kipnis.
Since the text in the Bible has been edited and re-edited many times, the team faced big challenges finding segments that retained their original wording and language.
Once found, these biblical texts were often very short—sometimes just a few verses—which made most standard statistical methods and traditional machine learning unsuitable for their analysis. They had to develop a custom approach that could handle such limited data.
Limited data often brings fears of inaccuracy. "We spent a lot of time convincing ourselves that the results we were getting weren't just garbage," said Faigenbaum-Golovin. "We had to be absolutely sure of the statistical significance."
To circumvent the issue, instead of using traditional machine learning, which requires lots of training data, the researchers used a simpler, more direct method. They compared sentence patterns and how often certain words or word roots (lemmas) appeared in different texts, to see if they were likely written by the same group of authors.
A surprising find? The team discovered that although the two sections of the Ark Narrative in the Books of Samuel address the same theme and are sometimes regarded as parts of a single narrative, the text in 1 Samuel does not align with any of the three corpora, whereas the chapter in 2 Samuel shows affinity with the Deuteronomistic History (Joshua to Kings).
Looking forward, Faigenbaum-Golovin said the same technique can be used for other historical documents. "If you're looking at document fragments to find out if they were written by Abraham Lincoln, for example, this method can help determine if they are real or just a forgery."
"The study introduces a new paradigm for analyzing ancient texts," summarized Finkelstein.
Faigenbaum-Golovin and her team are now looking at using the same methodology to unearth new discoveries about other ancient texts, like the Dead Sea Scrolls. She emphasized how much she enjoyed the long-term cross-disciplinary partnership.
"It's such a unique collaboration between science and the humanities," she said. "It's a surprising symbiosis, and I'm lucky to work with people who use innovative research to push boundaries."
More information: Shira Faigenbaum-Golovin et al, Critical biblical studies via word frequency analysis: Unveiling text authorship, PLOS One (2025). DOI: 10.1371/journal.pone.0322905
Journal information: PLoS ONE"
https://phys.org/news/2025-06-ai-reveals-hidden-language-patterns.html
#metaglossia_mundus
Depuis 25 ans, le programme Goldschmidt a formé près de 250 traducteurs et traductrices, contribuant à structurer un réseau international au service des échanges littéraires entre espaces francophones et germanophones...
"Goldschmidt : 25 ans au service des jeunes traducteurs littéraires
Jeudi 12 juin 2025, au Goethe-Institut de Paris, le programme Georges-Arthur Goldschmidt fête son quart de siècle. Créé en 2000 pour accompagner les jeunes traducteurs littéraires, ce dispositif, désormais franco-germano-suisse et, pour la première fois cette année, autrichien, marque un tournant avec une soirée dédiée à la traduction, à ses défis contemporains et à ses figures marquantes.
Le 05/06/2025 à 13:20 par Dépêche
Soutenu par l’OFAJ, France Livre, la Foire de Francfort, Pro Helvetia et le ministère autrichien des Affaires européennes et internationales, l’événement se veut à la fois professionnel, réflexif et convivial.
Le programme s'ouvrira à 17h30 avec une table ronde réunissant trois traducteurs issus de différentes générations du programme : Andreas Jandl, Régis Quatresous et Lionel Felchlin. Modérée par Hannah Sandvoss (France Livre), la discussion portera sur l’insertion professionnelle, les pratiques collectives de traduction et l’impact croissant de l’intelligence artificielle sur le métier.
À 18h45, l’auteur allemand Ulrich Effenhauser dialoguera avec Carole Fily, sa traductrice française formée au programme en 2011. Ensemble, ils reviendront sur leur collaboration autour du roman Le Fantôme de Mexico, un polar historique publié en février 2025 chez Actes Sud, où l'intrigue mêle espionnage, guerre froide et catastrophe de Tchernobyl.
À 19h45, Justine Coquel et Valentin Decoppet animeront un atelier original de « craduction » : une initiation ludique à la traduction sans connaissance préalable de la langue source, mêlant invention collective et écriture créative. Ouvert à tous, cet exercice interroge les frontières entre traduire, interpréter et écrire.
À LIRE - Dans les marges de Kafka : un manuscrit rare vendu à prix d’or
La soirée se conclura à partir de 20h30 par un moment festif, propice aux échanges informels entre professionnels, anciens participants et partenaires institutionnels.
Depuis 25 ans, le programme Goldschmidt a formé près de 250 traducteurs et traductrices, contribuant à structurer un réseau international au service des échanges littéraires entre espaces francophones et germanophones.
Crédits illustration : Programme Georges-Arthur Goldschmidt
Par Dépêche
Contact : depeche@actualitte.com
https://actualitte.com/article/124252/rencontres-dedicaces/goldschmidt-25-ans-au-service-des-jeunes-traducteurs-litteraires
#metaglossia_mundus
Nous sommes allés à la rencontre de Marine Borel à l’occasion de la remise du grand prix de linguistique et de philologie 2024 de l’Académie royale de langue et de littérature française de Belgique en mars 2025 pour son livre "Les formes verbales surcomposées en français".
"[Interview] Marine Borel, récipiendaire du grand prix de linguistique et de philologie 2024 de l’Académie royale de langue et de littérature française de Belgique
Publié le 4/06/2025
Nous sommes allés à la rencontre de Marine Borel à l’occasion de la remise du grand prix de linguistique et de philologie 2024 de l’Académie royale de langue et de littérature française de Belgique en mars 2025 pour son livre Les formes verbales surcomposées en français.
Tout d’abord, toutes nos félicitations ! Pouvez-vous nous en dire plus sur votre travail ?
Marine Borel : Cet ouvrage est la version retravaillée de ma thèse de doctorat, qui portait, comme l’indique son titre, sur les formes verbales dites « surcomposées » en français. Il s’agit des formes verbales qui comportent un élément auxiliaire de plus que les formes composées – comme par exemple j’ai eu fait ou j’avais eu fait. Dans mon livre, je montre qu’il existe en fait deux paradigmes distincts de formes surcomposées. Pour illustrer cela, prenons l’exemple du passé surcomposé. Il existe d’une part un passé surcomposé dit « standard », qui est attesté dans l’ensemble du monde francophone, qui se trouve le plus souvent dans des subordonnées temporelles et qui peut être considéré comme un homologue du passé antérieur. Par exemple : « Lorsqu’elle a eu mangé, elle est allée se coucher. » Et il existe d’autre part un passé surcomposé dit « régional », qui n’est utilisé que dans les régions où on parlait autrefois des dialectes occitans ou francoprovençaux et qui a une valeur qu’en linguistique on appelle « expérientielle » – ce qui signifie que ces formes régionales peuvent toujours être glosées par la formule ‘il est déjà arrivé que’. Par exemple, l’énoncé (authentique) « J’ai eu mangé de la marmotte à un anniversaire » signifie ‘il m’est déjà arrivé de manger de la marmotte à un anniversaire’.
Ces formes « régionales » semblent plutôt exotiques à des oreilles lorraines. Mais elles sont, comme vous le dites, bien attestées dans le Sud de la France et dans le domaine francoprovençal, dont fait partie la Suisse romande, d’où vous êtes originaire.
MB : Oui, même si ces formes étonnent celles et ceux qui ne les emploient pas, la région dans laquelle elles sont utilisées est loin d’être marginale : le domaine d’oc couvre tout le tiers Sud de la France, dans lequel se trouvent notamment les villes de Marseille, de Toulouse, de Nice, de Montpellier, de Bordeaux, de Perpignan et de Limoges ; quant au domaine francoprovençal, il correspond à la zone, à l’Est, dans laquelle se trouvent Lyon, Saint-Étienne et Grenoble et, bien sûr, la Suisse romande – où elles sont particulièrement vivaces. En Romandie, on les entend tous les jours, non seulement avec l’auxiliaire « avoir » mais aussi avec l’auxiliaire « être ». Et pour celles et ceux qui ne les utilisent pas (comme les Lorrain·e·s), les formes avec « être » semblent encore plus exotiques que celles avec « avoir », puisqu’elles se construisent sur le modèle de je suis eu parti·e – et non sur celui de j’ai été parti·e comme c’est le cas des formes dites standard. Voici deux exemples authentiques recueillis en Suisse romande : « Il est eu venu boire l’apéro chez moi » et « Tu es eu allée à Nancy pour le voir ? ». Ce dernier exemple a été produit par une amie qui me demandait s’il m’était déjà arrivé d’aller à Nancy pour rencontrer celui qui était à l’époque mon directeur de thèse, Denis Apothéloz.
Pouvez-vous nous en dire plus votre parcours et sur ce qui vous a amenée à Nancy ?
MB : Je suis née en 1987, j’ai grandi en Suisse romande, dans le canton de Neuchâtel, puis j’ai fait des études de langue et de littérature françaises à l’Université de Fribourg. Après mes études, j’ai décidé de faire une thèse de doctorat en linguistique française et, alors que je réfléchissais à un sujet de recherches, je suis tombée sur la thèse d’un linguiste suisse, Maurice Cornu, publiée en 1953, qui portait précisément sur les formes verbales surcomposées. J’ai lu cette thèse d’une seule traite et j’ai tout de suite su que je voulais consacrer mes recherches à ce sujet. J’ai aussitôt lu tout ce qui était à disposition sur cette question. C’est ainsi que j’ai découvert les travaux de Denis Apothéloz, également suisse d’origine, professeur de linguistique (aujourd’hui émérite) à l’Université de Lorraine et membre de l’ATILF, des travaux qui ont marqué un pas décisif dans la compréhension des formes surcomposées régionales. Je l’ai donc contacté pour lui demander s’il acceptait de co-diriger ma thèse dans le cadre d’une cotutelle de thèse entre l’Université de Fribourg et l’Université de Lorraine. Et il s’est tout de suite montré enthousiaste.
C’est donc dans ce cadre que vous êtes venue passer une année à Nancy et que vous avez été rattachée comme doctorante à l’ATILF ?
MB : Oui, j’ai eu la chance d’obtenir deux soutiens financiers qui m’ont permis de venir régulièrement à Nancy et d’y passer une année : la bourse de mobilité pour doctorant·e·s du Fonds national suisse de la recherche scientifique (FNS) et une contribution financière à la cotutelle de thèse de la conférence des recteurs des universités suisses (CRUS). Lors de mon séjour à Nancy, en 2016, j’ai été rattachée au laboratoire ATILF, où j’ai été extrêmement bien accueillie. Je garde notamment de très bons souvenirs des réunions de l’équipe de recherche « Discours, langue et cognition » à laquelle j’étais intégrée et des conférences qui étaient régulièrement organisées. J’ai beaucoup appris durant cette année et, surtout, j’ai pu avancer dans mes recherches dans un cadre idéal. Depuis, je suis bien sûr restée en contact non seulement avec le professeur Denis Apothéloz mais également avec les anciens collègues de l’ATILF – je me réjouis d’ailleurs beaucoup de ma participation à deux journées d’études organisées en novembre prochain par le professeur Yvon Keromnes, qui porteront précisément sur les formes verbales surcomposées.
Que faites-vous aujourd’hui ? Où travaillez-vous ?
MB : Aujourd’hui, j’enseigne la linguistique française, en tant que lectrice à l’Université de Zurich et en tant que chargée de cours dans les universités de Fribourg et de Bâle. Je travaille également comme responsable de projets au Forum Helveticum, une association qui s’engage pour les échanges et la cohésion entre les quatre régions linguistiques de la Suisse.
Où pouvons-nous trouver votre ouvrage ?
MB : Mon livre est disponible en libre accès sur le site de la maison d’édition Peter Lang | 2024 | Monographies | 622 pages"
https://factuel.univ-lorraine.fr/node/30288
#metaglossia_mundus
"The 3rd Spring School in Translation Studies Conference: "Translation & Imagination"
The third Lisbon Spring School in Translation Studies Conference will be held in early June at the Universidade Católica Portuguesa in Lisbon, Portugal. This year's Spring school, with the title 'Translation and the Imagination', will explore the multiple ways in which imagination weaves itself into the interpretation and the textual fabric of different forms of translation, from age-old forms to the new challenges posed by intermedial and transmedial texts, and computer-assisted translation. It will also look into the part imagination plays in translation and how this is affected by the emergence of AI translation.
Dr. Delphine Grass from the School of Global Affairs will deliver a paper entitled 'A Particular Kind of Relation: Translation, Praxis and the Social Imagination’. Her paper explores creative-critical translation practices which have engaged a critique of the politics of translation by deviating, slowing down or mattering translational relations in visual and textural terms to side-step oppressive rules of international and ecological engagement. It will also delve into the relationship between creativity and translation in the era of GenAI from the critical perspective of creative-critical translation practices.
Alongside its new MA Global Leadership programme, the School of Global Affairs also offers an established MA Translation Studies degree which allows students to develop their linguistic and technical skills to move into a wide range of careers in translation."
https://www.lancaster.ac.uk/global-affairs/news/lancaster-researcher-to-keynote-international-translation-studies-conference
#metaglossia_mundus
The promise echoed in diplomatic halls and joint venture signings embedded in a partnership forged in "mutual respect" (互相尊重) and "win-win cooperation" (共同利益) seems to have lost its true definition
Translation must breed respectful relationships
Translator's lament: When words bridge worlds, but respect falters
Dr Lucy Anning
Jun - 04 - 2025 , 09:57
The promise echoed in diplomatic halls and joint venture signings embedded in a partnership forged in "mutual respect" (互相尊重) and "win-win cooperation" (共同利益) seems to have lost its true definition.
For Chinese-to-English translators, particularly those in Ghana, these are not mere slogans but the very foundation upon which our professional existence is supposed to rest.
Translators are the linguistical bridges, the cultural interpreters and the vital conduits facilitating the vast economic engagement between China and Africa.
Yet, for too many of us, the reality behind the boardroom doors and factory floors is a soul-eroding indignity that starkly contradicts the official rhetoric.
This is the hidden ordeal of the translator, caught between nations and battling for basic human dignity.
The challenges within the ecosystem of Chinese companies operating in Ghana are not merely professional hurdles; they are profound assaults on personhood.
The core wound, as articulated with raw pain by one translator, is the pervasive sense of being viewed as "lower than human beings."
This dehumanisation manifests in chillingly tangible ways.
Many translators have reported salaries that are an insult to their skilled labour, a sheer pittance, "nothing to write home about."
Beyond the financial disrespect lies a spectrum of mistreatment consisting of routine bullying, verbal abuse that stings in any language, blatant infractions of Ghanaian labour laws and even alleged physical intimidation.
The workplace becomes a pressure cooker of injustice for these staff.
Rule of law, FDI initiatives?
The dissonance deepens when considering the broader context. These companies operate in the Ghanaian economy, extracting significant profit, often amidst accusations of tax avoidance and aggressive acquisition of local assets.
As one victimised voice starkly put it, "They are buying lands and properties like crazy. In tandem, for Zhongguo (Chinese economy), no foreigner has these kinds of privileges there."
This economic imbalance fuels the bitter question: Why must Ghanaians (and other African economies), particularly those facilitating this very enterprise, endure such degradation within their own homeland?
The principle seems grotesquely one-sided, in that "Chinese can mistreat African people within their jurisdiction, thus home country.
However, Africans can't take this abuse within their jurisdictive home countries... and allow these foreigners to go scot-free."
Corrupts leaders
The litany of grievances extends to unethical practices encapsulated in bribes, illegal dealings and opaque connections, creating an environment where workers, translators included, feel treated not as partners or even employees, but as expendable machines.
This is not just about poor working conditions; it’s a fundamental betrayal of the translator’s purpose. Our mastery of Putonghua is not merely a technical skill to facilitate transactions; it is the power to communicate as equals.
"It is also to be able to address them, speak to them like equal men, for them to hear us in a language that they understand," the source implores.
Yet, the language of respect we translate for official pronouncements rarely permeates daily interactions.
The grand narrative of mutual benefit rings hollow when met with daily doses of contempt.
Mitigating mechanisms
Mitigating this deep-seated crisis demands concerted, multi-level action.
Firstly, translators must organise. Collective voices are harder to ignore.
Forming professional associations or leveraging existing unions provides a platform to share experiences, document abuses systematically and advocate collectively for fair contracts, decent wages and dignified treatment.
Secondly, legal empowerment is non-negotiable. Translators need accessible legal resources and unwavering support to challenge labour law violations, unsafe conditions and illegal practices.
Reporting channels must be strengthened and protected.
Thirdly, Ghanaian (and other African countries) authorities must enforce their sovereignty.
This means rigorous application and monitoring of labour laws, tax regulations and investment codes pertaining to foreign businesses.
Permissiveness towards exploitation because of investment is a Faustian bargain that erodes national dignity and long-term stability.
Diplomatic channels must also be utilised to consistently raise these specific worker grievances, moving beyond abstract cooperation principles.
Finally, translators must wield their unique power, thus using their linguistic skills not just to transmit, but to assert.
Clearly articulating boundaries, expectations of respect and legal rights in the language understood is a form of resistance and empowerment.
Conclusion
From whispered grievances to a roar for respect, the whispered frustrations in translation booths and on factory floors are coalescing into an undeniable roar.
The ordeal of the Chinese-to-English translator in Ghana is a microcosm of a partnership dangerously out of balance.
It exposes the chasm between the polished language of diplomacy and the raw language of daily indignity.
Ghanaian translators are not mere linguistic tools; they are Ghanaian professionals, deserving fundamental respect on their own soil.
To ignore their suffering, to tolerate the blatant hypocrisy where "mutual respect" is preached but subjugation is practised, is to undermine the very foundation of the Ghana-China relationship itself.
The path forward isn't a revolution born of desperation, but a resolute demand for the respect already promised.
It’s time for Ghana to ensure its laws protect its people, for companies to uphold their professed values and for translators to cease being the unheard interpreters of their own oppression. Enough is indeed enough.
The bridges we build must be walked upon with dignity by all, or they will crumble under the weight of hypocrisy.
The next words translated must herald genuine change, or they will be the last before the silence of defiance falls.
The writer is with the DevAfrica Institute,
Accra, Ghana.
E-mail: Lucyann@
devafricainstitute.org"
https://www.graphic.com.gh/features/opinion/ghana-news-translators-lament-when-words-bridge-worlds-but-respect-falters.html
#metaglossia_mundus
"British translator Karen Leeder and German writer Durs Grünbein win Griffin Poetry Prize
Nicole Thompson, The Canadian Press
TORONTO — As German writer Durs Grünbein accepted the Griffin Poetry Prize at a ceremony in Toronto on Wednesday night, he pointed to the Ukrainian flag ribbon pinned to his lapel and launched into an explanation.
"It's not only for Ukraine. This is now a war on Europe. If we do not stand together, one day, this continent will vanish," he said of Russia's invasion of Ukraine.
"I hope poetry is a support to ideas of democracy, humanity ... That's why I'm writing poetry."
The Berlin-based Grünbein shares the $130,000 award with British scholar Karen Leeder, a professor of German language and literature at Oxford University, who translated "Psyche Running: Selected Poems, 2005-2022" into English.
The Griffin judges praised the collection as being "universal, lyrical, philosophical."
Leeder, who as the translator receives 60 per cent of the prize money, said she was shocked to hear her name called. In fact, she hasn't thought of what she'll do with the winnings.
"We were onstage with such amazing poets, all of them," she said of the other finalists.
"I was really struck by how each poet or translator-poet read a different kind of extraordinary poetry at the highest level. They were all so different."
For his part, Grünbein said he was humbled by the experience — and didn't know how much money came with the prize.
This is the second time he has been up for the award.
"Ashes for Breakfast: Selected Poems," translated by Michael Hoffmann, was a finalist for the International Griffin Poetry Prize in 2006, before the Canadian and global prizes were combined.
Margaret Atwood also received the $25,000 Lifetime Recognition Award at Wednesday night's ceremony.
In his acceptance speech, Grünbein said he hadn't realized she was such a skilled poet.
That changed when he read two of her poems on Wednesday night, he said.
"I realized immediately that these lines were so concentrated, focused on — in that case — very important historical, political things," he said.
Earlier in the evening, Atwood received a standing ovation as she walked onstage.
Though she’s best known for novels including “The Handmaid’s Tale,” Atwood got her start as a poet. She’s published more than a dozen collections since the 1960s.
She’s also a founding trustee of the Griffin Poetry Prize, which was first handed out in 2000.
“We did have trustee meetings and they consisted of (prize founder Scott Griffin) telling us what to do, and us making suggestions that were rejected,” Atwood recounted.
Whitehorse poet Dawn Macdonald also read from her debut collection “Northerny," which won the $10,000 Canadian First Book Prize.
“Margaret Atwood is a bit of a tough act to follow,” she quipped before reciting a poem.
The other finalists for the Griffin were "The Great Zoo," translated by Aaron Coleman from Nicolas Guillen's original Spanish; “Kiss the Eyes of Peace," translated by Brian Henry from the original Slovenian by Tomaž Šalamun; Carl Phillips for "Scattered Snows, to the North”; and Diane Seuss for "Modern Poetry."
Each of the finalists receives $10,000.
This report by The Canadian Press was first published June 4, 2025.
Nicole Thompson, The Canadian Press"
https://www.westernwheel.ca/lifestyle/british-translator-karen-leeder-and-german-writer-durs-grunbein-win-griffin-poetry-prize-10756993
#metaglossia_mundus
Left and right brains hear speech differently, yet how this divide forms was unclear − until mouse studies showed each hemisphere runs on its own developmental clock.
Published: June 4, 2025 2.45pm SAST
Hysell V. Oviedo, Washington University in St. Louis
Some of the most complex cognitive functions are possible because different sides of your brain control them. Chief among them is speech perception, the ability to interpret language. In people, the speech perception process is typically dominated by the left hemisphere.
Your brain breaks apart fleeting streams of acoustic information into parallel channels – linguistic, emotional and musical – and acts as a biological multicore processor. Although scientists have recognized this division of cognitive labor for over 160 years, the mechanisms underpinning it remain poorly understood.
Researchers know that distinct subgroups of neurons must be tuned to different frequencies and timing of sound. In recent decades, studies on animal models, especially in rodents, have confirmed that splitting sound processing across the brain is not uniquely human, opening the door to more closely dissecting how this occurs.
Yet a central puzzle persists: What makes near-identical regions in opposite hemispheres of the brain process different types of information?
Answering that question promises broader insight into how experience sculpts neural circuits during critical periods of early development, and why that process is disrupted in neurodevelopmental disorders.
Timing is everything
Sensory processing of sounds begins in the cochlea, a part of the inner ear where sound frequencies are converted into electricity and forwarded to the auditory cortex of the brain. Researchers believe that the division of labor across brain hemispheres required to recognize sound patterns begins in this region.
For more than a decade, my work as a neuroscientist has focused on the auditory cortex. My lab has shown that mice process sound differently in the left and right hemispheres of their brains, and we have worked to tease apart the underlying circuitry.
For example, we’ve found the left side of the brain has more focused, specialized connections that may help detect key features of speech, such as distinguishing one word from another. Meanwhile, the right side is more broadly connected, suited for processing melodies and the intonation of speech.
Sound information moves through the cochlea to the brain. Jonathan E. Peelle, CC BY-SA
We tackled the question of how these left-right differences in hearing develop in our latest work, and our results underscore the adage that timing is everything.
We tracked how neural circuits in the left and right auditory cortex develop from early life to adulthood. To do this, we recorded electrical signals in mouse brains to observe how the auditory cortex matures and to see how sound experiences shape its structure.
Surprisingly, we found that the right hemisphere consistently outpaced the left in development, showing more rapid growth and refinement. This suggests there are critical windows of development – brief periods when the brain is especially adaptive and sensitive to environmental sound – specific to each hemisphere that occur at different times.
To test the consequences of this asynchrony, we exposed young mice to specific tones during these sensitive periods. In adulthood, we found that where sound is processed in their brains was permanently skewed. Animals that heard tones during the right hemisphere’s earlier critical window had an overrepresentation of those frequencies mapped in the right auditory cortex.
Adding yet another layer of complexity, we found that these critical windows vary by sex. The right hemisphere critical window opens earlier in female mice, and the left hemisphere window opens just days later. In contrast, male mice had a very sensitive right hemisphere critical window, but no detectable window on the left. This points to the elusive role sex may play in brain plasticity.
Our findings provide a new way to understand how different hemispheres of the brain process sound and why this might vary for different people. They also provide evidence that parallel areas of the brain are not interchangeable: the brain can encode the same sound in radically different ways, depending on when it occurs and which hemisphere is primed to receive it.
Speech and neurodevelopment
The division of labor between brain hemispheres is a hallmark of many human cognitive functions, especially language. This is often disrupted in neuropsychiatric conditions such as autism and schizophrenia.
Reduced language information encoding in the left hemisphere is a strong indication of auditory hallucinations in schizophrenia. And a shift from left- to right-hemisphere language processing is characteristic of autism, where language development is often impaired.
Children with certain neurodevelopmental conditions may have trouble processing speech. Towfiqu Ahamed/iStock via Getty Images Plus
Strikingly, the right hemisphere of people with autism seems to respond earlier to sound than the left hemisphere, echoing the accelerated right-side maturation we saw in our study on mice. Our findings suggest that this early dominance of the right hemisphere in encoding sound information might amplify its control of auditory processing, deepening the imbalance between hemispheres.
These insights deepen our understanding of how language-related areas in the brain typically develop and can help scientists design earlier and more targeted treatments to support early speech, especially for children with neurodevelopmental language disorders."
https://theconversation.com/your-left-and-right-brain-hear-language-differently-a-neuroscientist-explains-how-257436
#metaglossia_mundus
"...In ancient India, the most powerful god was known as “sky father,” or in the Sanskrit language, Dyaus pita. Sound it out. Can you see where this is going? In Greece, his equivalent was Zeus pater; in Rome, Jupiter. English speakers have always been used to tracing the etymologies of their words back to the classical languages of Europe, but the suggestion, in the late 18th century, that there were also clear and consistent features in common with languages spoken much farther to the east was an astonishing and exhilarating one.
If we think of languages as belonging to families, this is like finding out that an acquaintance is really a distant cousin: It points to a shared ancestor, a unity – and a divergence – somewhere in the past. What would it mean for those cultural origin stories – for Who We Are – to speak not of European but of Indo-European? “Of what ancient and fantastic encounters were these the fading echoes?” asks Laura Spinney in her latest book, “Proto: How One Ancient Language Went Global.”
There are about 7,000 languages spoken in the world today; they can be divided into about 140 families. Nevertheless, the languages most of us speak belong to just five: Indo-European, Sino-Tibetan, Niger-Congo, Afro-Asiatic and Austronesian. And of these, Indo-European and Sino-Tibetan (which includes Mandarin) are the biggest. As Spinney notes, “Almost every second person on Earth speaks Indo-European.”
- ADVERTISEMENT -
Such a large family is, naturally, a diverse one. Among the Indo-European languages are English, German, Spanish, Italian, Russian, Hindi-Urdu and Persian. At the head of the family tree is the postulated language we call Proto-Indo-European, or PIE for short. It was spoken – but not written down, as it emerged in the days before writing – some five or six millennia ago, somewhere around the Black Sea. (Exactly where is an open question. We’ll come to that later.)
Although we have no documentation of what PIE sounded like, it is possible to triangulate backward using the languages we do know. Over time, the way we speak – the set of sounds we use – gradually changes. This is not a chaotic, willy-nilly drift; rather, the shifts obey a set of general rules. To give you an example, compare the group of words where Latin has a p sound but the English equivalent has f: pater/father; pisces/fish, and so on. We can posit, then, that at an early point in their development, Germanic languages went down an f path from their PIE root, while other languages stayed with p.
Using these sound-change rules to identify ancestral similarities among languages, we can reconstruct a corpus of about 1,600 PIE words, including (with apologies for the somewhat rebarbative notation) family relations like *méh₂tēr and *bʰréh₂tēr and numbers like *duó-, *tréy. If you were living on the eastern fringes of Europe some five millennia ago, you might stay around your *domo and drink some *h₂melǵ or maybe ride out on your *éḱwos to lie under the night sky and fix your eye on a *h₂ster.
So this is the original material. But how, to use Spinney’s subtitle, did this ancient language go global? Most of “Proto” is spent picking through the genetic and archaeological evidence to explain how PIE, or its descendants, came to be spoken from Galway to Kolkata.
Whatever its contested origins, by the third millennium B.C., PIE was being spoken in the Pontic steppe, the region of grasslands to the north of the Black Sea, by a culture known as the Yamnaya, who were nomadic and had ox-drawn wagons as well as, probably, domesticated horses. It was the Yamnaya – whose migration routes were studded with large burial mounds, still discernible, that they built into the landscape – who carried the language deep into Asia. Later branches of the tree, each with its own history, include the Baltic and Slavic languages, not to mention Tocharian, Albanian, Armenian, Greek, Celtic, Germanic and Indo-Iranian.
“Proto,” it has to be said, is a dense read. Spinney’s approach is a thorough one, so that the chapters tack back and forth in time as we follow each of these branches in turn. It is easy to lose sight of the timeline while concentrating on the differences between Hittites, Hattians and Hurrians, between Lydians and Luwians. Spinney’s background is as a science journalist, and it shows in the way she has assimilated a vast amount of up-to-date scholarly information and presented it concisely and accurately. Nevertheless, this is still a knotty story.
Spinney leavens the history lesson with sections dealing with the present: pen portraits of academics she has interviewed in the archaeological trenches of Ukraine and Russia as well as on the more rarefied battlefields of international academic conferences. As an observer rather than a participant, Spinney is comfortable laying out different sides of some of the ongoing debates, the biggest of all being where exactly the ancestral language was first spoken. Did PIE arise in Anatolia (roughly modern-day Turkey) before spreading first west into Europe along with farming and then east again with the Yamnaya? Or did it originate in the steppe, borne west by marauding nomads who replaced the populations they drove out? Or – the latest linguistic hypothesis – were these Anatolian and steppe languages both offspring of an even older ancestor spoken in the northern Caucasus (now southern Russia) some 6,000 years ago?
“Proto” is an impressive piece of work – hard but ultimately rewarding. And if the debates Spinney has delved into are still far from being resolved, time at least is an abundant resource. One Bulgarian archaeologist who spoke to Spinney addresses the ebb and flow of academic funding with an admirable lack of ego: “These people have lain in the ground for thousands of years. Who cares if it’s me or someone else who digs them up.”
– – –
Dennis Duncan teaches English literature at University College London...
By Dennis Duncan
June 5, 2025
https://www.newsindiatimes.com/our-languages-have-more-in-common-than-you-might-think/
#metaglossia_mundus
Editors spend hours cutting clips, adding text, fixing sound, and more. Now, smart editors are changing the way they work. Instead of spending so much time editing, they are using smart tools to help them finish faster. This gives them more time to focus on the fun and creative parts—like making stories, adding humour, or talking to their audience.
"Why smart editors are spending less time editing and more time creating
By Dimsumdaily Hong Kong -11:30PM Wed
4th June 2025 – (Hong Kong) In today’s world, video content is everywhere. From TikTok to YouTube, people watch short and long videos every day. But what many don’t know is that editing videos takes a lot of time. Editors spend hours cutting clips, adding text, fixing sound, and more. Now, smart editors are changing the way they work. Instead of spending so much time editing, they are using smart tools to help them finish faster. This gives them more time to focus on the fun and creative parts—like making stories, adding humour, or talking to their audience. One tool that many editors now use is CapCut’s text to speech. It helps them turn written words into a real-sounding voice. So, if you don’t want to use your own voice in a video, or if you are camera shy, you can just write what you want to say and use this tool. It will read it out loud in the video for you. This saves a lot of time and also makes your videos sound more professional.
The Old Way of Editing Was Too Slow
Before all these smart tools came, editors had to do everything manually. They needed to cut videos one by one, fix the sound, remove background noise, add music, and write subtitles—all by hand. This made the process very slow. Also, if you made a mistake, you had to go back and fix everything again. Even small changes would take hours. Editors had less time to think about ideas or try new things. Most of their time went into fixing videos instead of creating something new.
New Tools Are Helping Creators
But now, things are different. With new tools like AI and automation, editing has become much easier and faster. Editors don’t need to do everything by hand anymore. They can use smart software that edits videos for them, adds subtitles automatically, or even creates a full video based on a script. One great example is the CapCut’s AI video generator. It lets you create a full video just by writing some text. You can choose styles, music, and visuals, and the tool will build the video for you. This is super helpful for busy creators who want to post often but don’t have time to do heavy editing every day. By using tools like these, editors can now focus more on being creative. They can think about fun video ideas, talk to their followers, or plan their next big content piece. Editing becomes just a small step in the bigger process.
How Smart Editors Spend Their Time Now
Now that editing is easier, creators are doing more exciting things with their time. They spend more time planning, writing better scripts, and finding new trends online. Some are learning how to make videos that go viral, while others are building strong personal brands.They also test new video styles like animations, reactions, or storytelling shorts. Instead of fixing sound or editing frame-by-frame, they’re creating meaningful content. They collaborate with others and grow their communities. This shift has changed how people view video making. Today, even one person with a laptop and internet can build a huge audience if they use the right tools smartly.
How to Use Text to Speech in Your Videos (3 Easy Steps)
If you want to try it yourself, here is how you can use text to speech in your videos:
Step 1: Import your video
Go to the CapCut desktop video editor and click on the New Project > Import and upload your desired file.
Step 2: Convert text to speech
Go to Text > Add text and then type what you want to say in your video. Choose a voice from the options by clicking the “Text to Speech” tab. You can pick male or female voices, different accents, or even emotions like happy or serious.
Step 3: Export video
Once you’re happy with the result, export the video. You can now share it on TikTok, YouTube, Instagram, or wherever you like.
This process takes just a few minutes, but it gives your video a big boost in quality.
Smart Editors Use Voice Tools for a Creative Edge
Besides using text to speech, many smart creators are also using tools like CapCut’s voice changers. These let you change your voice in funny or dramatic ways. For example, you can sound like a robot, a cartoon character, or even an old man. Why is this useful? Because it makes your content more fun! People like to watch videos that are creative, unexpected, or entertaining. Using a voice changer can help you stand out from the crowd. It also helps if you’re doing storytelling or gaming content where different characters have different voices. And again—it saves time. You don’t need to act or use your real voice. The software does it all.
Why Less Editing Means Better Content
Some people might think that doing less editing is bad. But actually, it’s the opposite. When you spend less time editing, you have more time to do things that really matter.Here are a few benefits:
You can post more videos every week
You have time to test new ideas
You don’t get tired or stressed from long editing hours
You can grow your audience faster
You can make better content because you are focused on the message, not the edits
Final Thoughts
Video editing is changing. Smart editors are not spending long hours fixing small things anymore. Instead, they are using tools like text to speech, AI video generator, and voice changer to finish videos faster and better.If you’re a creator, student, teacher, or business owner—this is your chance. You don’t need to learn complicated software. Just use these tools and start creating! Be smart. Spend less time editing and more time sharing your message with the world."
https://www.dimsumdaily.hk/why-smart-editors-are-spending-less-time-editing-and-more-time-creating%EF%BF%BC/
#metaglossia_mundus
"These newsrooms desperately need the help these technologies provide, but they’re the ones being left out because they work in languages that are considered low-resource, so they are not a big priority for tech companies to support."
"These journalism pioneers are working to keep their countries’ languages alive in the age of AI news
“These newsrooms desperately need the help these technologies provide, but they’re the ones being left out because they work in languages that are considered low-resource, so they are not a big priority for tech companies to support.”
By GRETEL KAHN
JUNE 4, 2025, 10:28 A.M.
This story was originally published by the Reuters Institute for the Study of Journalism.
Since the launch of ChatGPT in 2022, newsrooms have been grappling with both the promise and the peril posed by generative AI. But not every publisher is equally prepared to pursue these opportunities. While newsrooms in the U.S. and Europe innovate and experiment with large language models (LLMs), many newsrooms in the Global South are being left behind.
While AI models in languages like English, Spanish, and French have ample training and resources, profound linguistic and cultural biases embedded within many mainstream AI tools pose substantial challenges for newsrooms and communities operating outside of dominant Western languages and cultural contexts.
For these newsrooms, gaps in data aren’t merely technical glitches but existential threats to their ability to participate equitably in this evolving digital ecosystem. What threats does this lack of data present to newsrooms in the Global South? How can these AI gaps be narrowed? To answer these questions, I spoke to six journalists and experts from India, the Philippines, Belarus, Nigeria, Paraguay, and Mali who are aiming to level the field.
What AI can (and can’t) do for low-resource languages
AI tools do not work well (or at all) for local, regional, and indigenous languages, according to all the sources I spoke to. This is particularly evident in countries where many of these languages are spoken beyond dominant ones like English, Spanish, or French. This language gap exacerbates inequalities for newsrooms and communities that operate in non-dominant languages.
Jaemark Tordecilla is a journalist, media advisor, and technologist from the Philippines focusing on AI and newsroom innovation. Having worked as a consultant for newsrooms across Asia, Tordecilla has seen a lot of curiosity among journalists about AI, but uptake has been mixed. For transcriptions — the AI functionality he has seen used the most — cost and language have often been an issue.
“For the longest time, [transcription] worked well for English and it didn’t work at all for Filipino, so journalists in the Philippines are only very recently getting to use these tools for their own reporting, but cost is an issue,” he said.
He described instances where journalists in the Philippines were forced to share accounts for paid subscriptions to transcription tools, which can create security issues. Things are even worse for journalists who use regional languages, for which these tools are essentially useless.
“The roll-out for support for regional languages has been slow and they are being left behind,” he said. “If your interview is in a regional language, then obviously you can’t use AI to process that. You can’t use AI to translate that into another language. You can’t use AI to say monitor town hall meetings in a regional language and make [them] more accessible to the public.”
Indian journalist Sannuta Raghu, who heads Scroll.in’s AI Lab and is now a journalist fellow at the Reuters Institute, has documented what these linguistic and cultural inequities look like in practice for newsrooms like hers.
AI tools don’t work very efficiently for most of the 22 official languages in India. Raghu listed issues like inaccurate outputs, hallucinations, and incorrect translations. A big issue, she said, is that, for a long time, AI tools were unable to account for nuances in language. Unlike English, for example, many Indian languages have large differences between spoken and written discourse.
“Asian countries speak in code-mixed language — for example, using multiple Hindi words and English words in a normal conversation,” she said, “which means we need rich enough data to be able to understand that.”
If there isn’t sufficient “good data” to train models on these specific languages and contexts, Raghu said, linguistic and cultural inequities are going to happen. Raghu attributed this lack of training data to a combination of complexity and lack of interest by Big Tech. But she also said the situation is starting to improve.
“Is it really a priority for you to optimize for all those languages in India from a tech product sales perspective? As you move eastward, the complexity of how societies use language changes. It becomes far more multilingual. It becomes far more complex. It becomes far more code-mixed. Those are the complexities that we live with, and that is not reflected in any of the models,” Raghu said.
AI ignores political biases
Beyond these inefficiencies, my sources also pointed out the cultural and political nuances that are missing from these models, which make them even more problematic for newsrooms.
For example, Raghu said that newsrooms are already noticing an American slant in everything AI generates. She described an instance where they were testing a tool to see how useful it’d be to help them write copy on cricket. For something as simple as explaining the sport, she says there were hallucinations, with players being made up, and the model simply not understanding the rules of the game.
“Up to 2.6 billion people follow cricket. It’s a huge cultural thing for us, Australia, Bangladesh, England…but the U.S. doesn’t play cricket, which is why a lot of the cultural aspects of this are not included in the models,” she said. “There is a lack of contextual training data. Cricket is very important for us, but we’re not able to do this with the LLMs because these models don’t quite understand the rules.”
Daria Minsky is a Belarusian media innovation specialist focusing on AI applications in journalism. After working with several newsrooms in exile, she has seen a lot of skepticism toward AI use — not because of factual errors, but because some of these models lack nuance in politically sensitive contexts.
Minsky used her own country as an example of how LLMs may simply repeat narratives put forth by authoritarian regimes.
“The word Belarusian is very politically loaded in terms of how you spell it, so I compared different models. ChatGPT actually spells the democratic or version of it, while DeepSeek uses the old, Soviet version of it,” she says. “It’s Belorussian versus Belarusian. Belorussian is this imperialistic term used by the regime. If newsrooms use Belorussian instead of Belarusian, they risk losing their audience and their reputation.”
AI models are trained based on data available online, which is why these models are more fine-tuned to English (and its contexts) than, say, Belarusian. Since the official narrative of authoritarian regimes is what is most available online, AI is being trained to follow that narrative.
These gaps in the training system have already been exploited by bad actors. A recent study by NewsGuard revealed that a Moscow-based disinformation network is deliberately infiltrating the retrieved data of AI chatbots, publishing false claims and propaganda for the purpose of influencing the responses of AI models on topics in the news. The result is more output that propagates propaganda and disinformation.
“I’ve heard of the same problems in Burma, for example, because the opposition uses Burma instead of Myanmar. I can see this type of problem even within the United States where I’m based, with the debate around using ‘Gulf of Mexico’ or ‘Gulf of America’ because Trump started to rename things just like in other dictatorial regimes,” she says.
How to close the gap
Despite these issues, some newsrooms in the Global South are taking the development of AI tools into their own hands.
Tama Media, a francophone West African news outlet, has launched Akili, an app, currently in beta, that provides fact-checking in local African languages through vocal interaction. Moïse Mounkoro, the editorial advisor in charge of Akili, traced the origin of the idea to two things: the fact that misinformation is rampant in West Africa and the realization that many people communicate via voice messages rather than through reading and writing.
“Educated people can read articles and they can fact-check,” Mounkoro said. “But for people who are illiterate or who don’t speak French, English or Spanish, the most common way to communicate is orally. I’m originally from Mali, and most of the conversations there are through WhatsApp voice messages. If you want to touch those people, you need to use their language.”
The Akili app uses AI to fact-check information by taking the question and finding an answer through its database of sources, which range from BBC Africa to Tama Media. The answer is then given orally to the user. To include more African languages like Wolof, Bambara, and Swahili, Akili’s team is experimenting with either using Google Translate’s API or building their own through online dictionaries.
“These AI technologies came from the West so they focus on their own languages. It’s like cameras: in the beginning, they were not made to photograph Black skin,” he says. “People need to be aware that they need to integrate other languages. Google has at least tried to make an effort by integrating many African languages these last couple years.”
In Paraguay digital news outlet El Surti is developing GuaraníAI. While the project is still in development, their goal is to build a chatbot to detect whether someone is speaking in this language and provide them with a response. In order to do that, they are developing a dataset of spoken Guaraní so that LLM engines can recognize oral speech in this Indigenous language, spoken by almost 12 million people.
Sebastián Auyanet, who leads on the project, told me they wanted to explore how those who don’t speak dominant languages are shut out from accessing LLMs. Guaraní is a language spoken still spoken widely throughout the Southern Cone, mainly in Paraguay, where it is an official language along with Spanish. Up to 90% of the non-indigenous population in Paraguay speaks Guaraní.
“Guaraní is an oral, unwritten language,” Auyanet said. “What we need is for any LLM engine to be able to recognize Guaraní speech and to be able to respond to these questions in Spanish. News consumption may be turning into ChatGPT, Perplexity, and other models. So there’s no way to get into that world if you speak in a language none of these systems can use.”
El Surti is organizing hackathons throughout Paraguay to test this dataset. Mozilla’s Common Voice is a platform designed to gather voice data for diverse languages and dialects to be used by LLMs. El Surti is using Common Voice to develop their minimum viable product, which aims to achieve 70% validation for spoken Guaraní within Mozilla’s datasets. With this degree of validation, the chatbot would be able to respond to queries in this Indigenous language.
A picture of the first hackathon of the project, which included conversations between El Surti’s journalists and Guaraní native speakers. (El Surti)
Ideally, Auyanet said, this project will allow El Surti to build an audience of Guaraní speakers who eventually will be able to interact with the outlet’s coverage by asking the chatbot questions in their own language.
“Right now we are excluding people who are only Guaraní speakers from the reporting that El Surti does,” he said. “This is an effort to bring them closer.”
In Nigeria, The Republic, a digital news outlet, is developing Minim, an AI-powered decentralized text-to-speech platform designed to support diverse languages and voices. They are actively training an AI model on specific African languages such as Nigerian Pidgin, Hausa, and Swahili, with plans to add more over time. The team is aiming to have a minimum viable product by the end of the year.
This model would allow independent creators to lend their own voices and train the AI in their unique vocal quality, including age, regional accents, and other demographics.
Editor-in-chief Wale Lawal told me that their goal is to engage audiences who speak these languages and to make their outlet more relevant to them. “We believe that global media has an indigenous language problem,” he said. “In a place like Africa you have a lot of people who are just automatically locked out of global media because of language.”
Minsky, the media consultant from Belarus, has been working with newsrooms in exile to develop an AI tool that allows for the automation of news monitoring from trusted sources. Her goal is to account for all the cultural and political nuances missing from the current models by allowing newsrooms to monitor very specific contextual sources, including local and hyper-local channels like Telegram.
This would include uploading archival data and historical data to fine-tune the output, and to explicitly prompt to control terminology and spelling (i.e. “don’t call Lukashenko president”).
No newsroom left behind
Newsrooms have been working without AI for decades before the first generative AI tools were released to the public. So what’s the problem if these tools are not available to everyone?
Tordecilla stressed that the gap between newsrooms is getting wider and pointed to things already happening: there are newsrooms in Manila who are doing AI investigations and newsrooms in rural Philippines who could benefit from the efficiencies that AI provides that are struggling for their survival.
“We are talking about newsrooms who have five people on their team, so every minute they spend on transcription is a minute they don’t spend reporting and editing stories,” Tordecilla said. “These newsrooms desperately need the help these technologies provide, but they’re the ones being left out because they work in languages that are considered low-resource, so they are not a big priority for tech companies to support.”
Indian journalist Sannuta Raghu said that addressing the AI digital gap is important for access, scale, and reaching a broader audience. Raghu mentioned a specific goal that her newsroom had when they started their AI lab: to build a multilingual text-to-video tool, as video content is extremely popular in India. With thousands of millions of Indians using smartphones and having access to very cheap internet, there is a significant opportunity to deliver journalism via video in these local languages. They have created Factivo 1.0, a URL-to-MP4 tool that creates accurate videos from news articles in a matter of minutes and which is available for major languages.
“For a small English-language newsroom like us, the universe available to us right now is about 30 million Indians who understand English,” she explains. “But to tap into the larger circle that is India, which is 1.4 billion people, we would need to be able to do language at scale.”"
https://www.niemanlab.org/2025/06/these-journalism-pioneers-are-working-to-keep-their-countries-languages-alive-in-the-age-of-ai-news/
#metaglossia_mundus
Tokenizer design significantly impacts language model performance,
yet evaluating tokenizer quality remains challenging.…
"Beyond Text Compression: Evaluating Tokenizers Across Scales
Authors Jonas F. Lotz†‡, António Vilarinho Lopes, Stephan Peitz, Hendra Setiawan, Leonardo Emili
Tokenizer design significantly impacts language model performance, yet evaluating tokenizer quality remains challenging. While text compression has emerged as a common intrinsic metric, recent work questions its reliability as a quality indicator. We investigate whether evaluating tokenizers on smaller models (350M parameters) reliably predicts their impact at larger scales (2.7B parameters). Through experiments with established tokenizers from widely-adopted language models, we find that tokenizer choice minimally affects English tasks but yields significant, scale-consistent differences in machine translation performance. Based on these findings, we propose additional intrinsic metrics that correlate more strongly with downstream performance than text compression. We combine these metrics into an evaluation framework that enables more reliable intrinsic tokenizer comparisons.
† Work done while at Apple
‡ University of Copenhagen & ROCKWOOL Foundation Research Unit
https://machinelearning.apple.com/research/beyond-text-compression
#metaglossia_mundus
"Development of intercultural communicative competence. Challenges and contributions for the training of health professionals
Authors
Juan Beltrán-Véliz Universidad de la Frontera
José Luis Gálvez-Nieto Universidad de la Frontera
Maura Klenner Loebel Universidad de la Frontera
Ana María Alarcón Universidad de la Frontera
Nathaly Vera-Gajardo Universidad Autónoma de Chile
Vol. 31 No. 1 (2025): NARRATIVAS, BIOÉTICA Y MEDICINA
Abstract
This paper reflects on intercultural communicative competence as a contribution to the training of professionals in health sciences. A reductionist, technocratic and culturally decontextualized health model is visualized, and the scarce development of "intercultural communicative competence" in health professionals is seen as a great barrier, which hinders communication with patients from other cultures. Therefore, it is vitally important to develop this competence in the training of these professionals. This will imply developing components that underlie this competence, that is, awareness, sensitivity and intercultural efficacy, which will allow the acquisition of cognitive, affective and behavioral skills that benefit effective and appropriate communication with patients from diverse cultural and social contexts. Likewise, they will allow them to understand in depth aspects, values and practices that underlie a culturally different health system, and will promote the advancement in the knowledge, understanding and development of innovative and relevant practices in the health system, from an intercultural and comprehensive approach. Finally, it will make it possible to overcome inequalities, misunderstandings, prejudices, exclusions and social injustices."
https://actabioethica.uchile.cl/index.php/AB/article/view/78367
#metaglossia_mundus
Scientists discover how story language evolves from scene-setting to action to resolution, analyzing thousands of fictional narratives.
"Scientists decode the universal language pattern in 40,000 stories
ELLSWORTH TOOHEY 8:46 AM WED JUN 4, 2025
"The Hare's Bride", a fairy tale from Buckow in Mecklenburg, by Tom Seidmann-Freud, (1924) Public Domain
A study analyzing 40,000 stories reveals what Aristotle intuited centuries ago — narratives follow a common, predictable structure consisting of three key elements.
Using computer analysis, researchers Ryan Boyd, Kate Blackburn, and James Pennebaker revealed how language patterns evolved across novels, short stories, romance novels, movies and other narratives. Their findings, published in Science in 2929, identified three consistent components that form narratives: At the beginning, articles and prepositions dominate as authors set the stage. The middle sees a rise in "cognitive tension words" like "think" and "realize" as characters grapple with conflicts. By the end, pronouns and auxiliary verbs increase as the action resolves.
This pattern held true regardless of genre, length, or even quality — highly-rated books and movies followed the same structure as poorly-rated ones. However, the researchers found that non-fiction formats like newspaper articles and TED talks deviated from this pattern, particularly in how they handle cognitive tension.
"We find that all different types of stories — novels, films, short stories, and even amateur writing — trend toward a universal narrative structure," the researchers write in their report. Rather than constraining creativity, they suggest this structure may provide an optimal way for humans to process and share information through stories.
They've created a website where anyone can analyze the narrative patterns of hundreds of published books and movie scripts, or upload their own texts for analysis."
https://boingboing.net/2025/06/04/scientists-decode-the-universal-language-pattern-in-40000-stories.html
#metaglossia_mundus
|
