 Your new post is loading...
|
Scooped by
Charles Tiayon
November 4, 2011 8:32 PM
|
Google is shutting down its free Translation API on 1 December, while Microsoft's remains free.
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
"AUDIO: Australian National Dictionary Centre could be shut down
NewsRadio
Pix: Dr Amanda Laugesen, the director of the Australian National Dictionary Centre at the Australian National University. (Supplied: ANU/Lannon Harley)
The Australian National Dictionary Centre conducts research into Australian English, and provides Oxford University Press with editorial expertise for their Australian dictionaries.
But now the centre could be shut down.
It's been earmarked as part of proposed cuts at the Australian National University."
https://www.abc.net.au/news/2025-07-15/naus_aussiedictionarynr_1507/105532588
#metaglossia_mundus
"Yes, it can turn French words into English ones, but true translation is so much more than that
Last week’s news about the launch of GlobeScribe, a publishing service offering AI translations of novels for the bargain price of $100 (£74), confirms what we’ve long suspected – that AI is being discussed not merely as a tool that can be useful for professionals, but as something to replace us entirely. As a human literary translator, I’m among those facing down that challenge. But we’re not replaceable just yet.
One line of dialogue in the story I’m translating this weekend reads, in its entirety: “T'es fatigué, toi?” Not the most taxing piece of translation work I’ve ever undertaken. I f you don’t know French, it’s basically “Are you tired?” – that’s the gist. But to give you more colour: the French is very casual, and uses an informal/intimate mode of address (tu rather than vous). Also encoded within this short phrase is the fact that the addressee is male; we’re asking if he’s fatigué, not if she’s fatiguée.
I am also a ‘large language model’ (God help me) – but I am also more than that..
“Are you tired?” doesn’t automatically tell us which other character is being addressed and nor does that English “you” testify to anything about this relationship, which matters because only yesterday this couple were calling one another vous, so this tu indicates a relational shift. As the translator, I notice these things and decide whether or not they’re significant, and act on them if they are. I draw on my ability to read French and to write English, and an ear for real-life dialogue, but also on my understanding that these are the same people we met back on page 16 (though not named), why the casualness is worth conveying in this specific situation, and what it means to be in a human relationship experiencing this sort of shift.
“Are you tired?” is inoffensive. It’s also readable, natural and smooth (the usual tiresome measures of quality when talking about translation). “Are you tired?” is also less than its source. As a reader, you might care about the loss; you might not. I do – for one thing, I’m writing a book at the moment, so I’m rather hoping for more than “basically, yeah, that’s the gist” from my own translators.
I’ve looked at plenty of machine translations. At this moment, I’m not worried about what AI (including large-language models trained on my copyright work) can actually do. As a human, I too am a “large language model” (God help me) – but I am also more than that. No, what AI can do is not yet a threat to human translators and human readers of sophisticated, textured human writing. The threat is what some people think it can do."
Daniel Hahn
Saturday 12 July 2025
AI won’t replace us human literary translators just yet
https://observer.co.uk/news/opinion-and-ideas/article/ai-wont-replace-us-human-literary-translators-just-yet
#metaglossia_mundus
Taraba revenue chief moves to digitize endangered indigenous languages
July 15, 2025
In a bold step toward preserving Nigeria’s linguistic heritage, retired Brigadier General Jeremiah Faransa has partnered with Izesan Limited to digitise the endangered Wurkun and Jiba languages spoken in Taraba State.
General Faransa, who currently chairs the Taraba State Internal Revenue Service and leads the state’s Special Task Force on Illegal Mining and Deforestation, pledged his full support for the initiative during a meeting with the leadership of Izesan on Thursday, July 11.
The project, which aligns with Nigeria’s commitment to UNESCO’s International Decade of Indigenous Languages (IDIL), aims to preserve native languages by integrating them into digital education platforms, governance, and cultural discourse.
Founder and CEO of Izesan Limited, Anthony Osekhuemen Otaigbe, emphasized the urgency of the collaboration, describing it as a “necessary response to cultural endangerment.”
“Language is not merely a tool for communication; it is a living archive of our identity, values, and ecological wisdom,” Otaigbe said.
Under the agreement, Izesan will develop curriculum-based educational content, as well as mobile and web applications, to support the learning and usage of Wurkun and Jiba languages across Taraba communities and beyond.
The CEO praised General Faransa’s commitment to cultural and environmental protection, noting that his endorsement provides strategic credibility to the project.
“Having General Faransa’s backing is a huge boost to our efforts. His involvement underscores the fact that heritage preservation is not just emotional—it’s essential to development,” Otaigbe added.
With over 400 Nigerian languages considered endangered, according to the National Institute for Nigerian Languages, experts warn that urgent action is needed to prevent cultural extinction. Language preservation is increasingly viewed as critical to advancing policy inclusion, economic development, and environmental sustainability—areas where indigenous knowledge plays a vital role.
Izesan Limited, Nigeria’s leading edtech company focused on indigenous language learning, has built a reputation for reconnecting African youth with their roots through innovative digital tools. Its products are used both locally and across the diaspora.
The digitisation of Wurkun and Jiba marks a significant milestone in harnessing technology for cultural sustainability and inclusive development in Nigeria."
https://www.vanguardngr.com/2025/07/taraba-revenue-chief-moves-to-digitize-endangered-indigenous-languages/
#metaglossia_mundus
"Ces villages de Loire-Atlantique inaugurent leurs panneaux en breton
Pour valoriser leur patrimoine linguistique, leurs origines et faire partie de l’association Breizh 5/5, les villagesde Saillé et Careil (Loire-Atlantique) ont installé des panneaux de leur nom bretons à leurs entrées.
Ouest-France
Publié le 14/07/2025 à 10h00
Pour développer les actions en faveur de la langue bretonne, et conforter l’adhésion de deux villages guérandais à l’association Breizh 5/5, deux panneaux bilingues ont été inaugurés, mercredi.
Il s’agit de ceux de Careil et de Saillé, installés aux entrées des deux communes. Comme le souligne le maire de Guérande, Nicolas Criaud, cela marque une étape symbolique et forte de valoriser notre patrimoine linguistique. Cette initiative s’inscrit dans une démarche cohérente et progressive de mise en valeur des villages et lieux-dits de la commune en breton »..."
#metaglossia_mundus
https://www.ouest-france.fr/pays-de-la-loire/loire-atlantique/ces-villages-de-loire-atlantique-inaugurent-leurs-panneaux-en-breton-becfaa72-5fee-11f0-b649-a8b202832ea5
Today’s AI models struggle to operate in smaller languages like Cantonese and Vietnamese, which are still spoken by tens of millions of people.
"The world’s best AI models operate in English. Other languages—even major ones like Cantonese—risk falling further behind
BY CECILIA HULT
July 15, 2025 at 8:20 AM EDT
How do you translate “dim sum”? Many English speakers would find the question strange, knowing the term refers to the large array of small dishes that accompanies a Cantonese-style brunch—and so doesn’t need translation.
But words like “dim sum” are a challenge for developers like Jacky Chan, who launched a Cantonese large language model last year through his startup Votee. It might be obvious to a human translator what words are loanwords and which need direct translation. Yet it’s less intuitive for machines.
“It’s not natural enough,” Chan says. “When you see it, you know it’s not something a human writes.”
Translation troubles are part of a growing list of issues when today’s AI models, strongest in English and other major languages, try to work in an array of smaller tongues still spoken by tens of millions of people.
When AI “models encounter a word they don’t know or that doesn’t exist in another culture, they will simply make up a translation,” explains Aliya Bhatia, a senior policy analyst at the Center of Democracy & Technology, where she researches issues related to multilingual AI. “As a result, many machine-created datasets could feature mistranslations, words that no native speaker actually uses in a specific language.”
LLMs need data, and lots of it. Text from books, articles and websites is broken down into smaller word sequences to form a model’s training dataset. From this, LLMs learn how to predict the next word in a sequence, eventually generating text.
AI can now generate text remarkably well—at least, it can in English. In other languages, performance lags significantly. Roughly half of all web content is in English, meaning there’s no shortage of digital resources for LLMs to learn from. Many other languages do not enjoy this same abundance.
Low-resource languages
So-called low resource languages are those with limited online data. Endangered languages, no longer being passed down to younger generations, clearly fall into this category. But widely spoken languages like Cantonese, Vietnamese and Bahasa Indonesia are also considered low-resource.
One reason could be limited internet access, which would prevent the creation of digital content. Another could be government regulation, which might limit what’s available online. Indonesia, for example, can remove online content without offering a way to appeal decisions. The resulting self-censorship may mean that available data in some regional languages might not represent authentic local culture.
This resource gap leads to a performance gap: Non-English LLMs are more likely to produce gibberish or inaccurate answers. LLMs also struggle with languages that don’t use Latin script, the set of letters used in English, as well as those with tonal features that are hard to represent in writing or code.
Currently, the best-performing models work in English and, to a lesser extent, Mandarin Chinese. That reflects where the world’s biggest tech companies are based. But outside of San Francisco and Hangzhou, a legion of developers, large and small, are trying to make AI work for everyone.
PAID CONTENT
A major step forward in PCG Power’s globalization strategy
FROM PCG POWER
South Korean internet firm Naver has built an LLM, HyperCLOVA X, which it claims is trained on 6,500 times more Korean data than GPT-4. Naver is also working in markets like Saudi Arabia and Thailand in a bid to expand its business creating “sovereign AI,” or AI tailored to a specific country’s needs. “We focus on what companies and governments that want to use AI would want, and what needs Big Tech can’t fulfill,” CEO Choi Soo-Yeon told Fortune last year.
In Indonesia, telecom operator Indosat and tech startup Goto are collaborating to launch a 70 billion parameter LLM that operates in Bahasa Indonesia as well as five other local languages, including Javanese, Balinese, and Bataknese.
One hurdle is scale. The most powerful LLMs are massive, made up of billions of word sequences converted into variables known as parameters. OpenAI’s GPT-4 is estimated to have around 1.8 trillion parameters. DeepSeek’s R1 has 671 billion.
Non-English LLMs seriously struggle to achieve this kind of scale. The Southeast Asian Languages in One Model (SEA-LION) project has trained two models from scratch: One with 3 billion parameters and one with 7 billion, much smaller than leading English and Chinese models.
Chan, from Votee, faces these struggles when dealing with Cantonese, spoken by 85 million people across southern China and Hong Kong. Cantonese uses different grammar for formal writing compared to informal writing and speech. Available digital data is scarce and often low-quality.
Training on digitalized Cantonese texts is like “learning from a library with many books, but they have lots of typos, they are poorly translated, or they’re just plain wrong,” says Chan.
Without a comprehensive dataset, an LLM can’t produce complete results. Data for low-resource language often skews towards formal texts—legal documents, religious texts, or Wikipedia entries—since these are more likely to be digitized. This bias can distort an LLM’s tone, vocabulary and style, and limit its knowledge.
LLMs have no inherent sense of what is true, and so false or incomplete information will be reproduced as fact. A model trained solely on Vietnamese pop music might struggle to accurately answer questions on historical events, particularly those not related to Vietnam.
Translating English content
Turning English content into the target language is one way to supplement the otherwise-limited training data. As Chan explains, “we synthesize the data using AI so that we can have more data to do the training.”
But machine translation carries risk. It can miss linguistic nuance or cultural context. A Georgia Tech study of cultural bias in Arabic LLMs found that AI models trained on Arabic datasets still exhibited Western bias, such as referencing alcoholic beverages in Islamic religious contexts. It turned out that much of the pre-training data for these models came from web-crawled Arabic content that was machine-translated from English, allowing cultural values to sneak through.
In the long-term, AI-generated content might end up polluting low-resource languages datasets. Chan likens it to “a photocopy of a photocopy,” with each iteration degrading the quality. In 2024, Nature warned of “model collapse,” where AI-generated text could contaminate the training data for future LLMs, leading to worse performance.
The threat is even greater for low-resource languages. With less genuine content out there, AI-generated content could quickly end up making up a larger share of what’s online in a given language.
Large businesses are starting to realize the opportunities in building a non-English AI. But while these companies are key players in their respective tech sectors, they’re still much smaller than giants like Alibaba, OpenAI, and Microsoft.
Bhatia says more organizations—both for-profit and not-for-profit—need to invest in multilingual AI if this new technology is to be truly global.
“If LLMs are going to be used to equip people with access to economic opportunities, educational resources, and more, they should work in the languages people use,” she says.
Fortune is bringing Brainstorm AI back to Asia on July 22-23 with the latest edition of our Brainstorm AI Singapore conference. Fortune will be convening the smartest people we know—technologists, entrepreneurs, Fortune Global 500 executives, investors, policymakers, and the brilliant minds in between—to explore and interrogate the most pressing questions about AI. Register here!
About the Author
CECILIA HULT
Cecilia Hult is an editorial intern based in Hong Kong.
https://fortune.com/asia/2025/07/15/ai-llm-language-english-cantonese-vietnamese-translation/
#metaglossia_mundus
"Pennsylvania's refugees speak well over 50 different languages. Those languages include:
Albanian
Amharic
Arabic
Armenian
Bassa
Bosnian
Bulgarian
Burmese
Cantonese
Chin
Creole
Croatian
Dinka
French
Gio
Hmong
Karen
Khmer
Kiswahili
Kpelle
Krahn
Krio
Kru
Kurdish
Lao
Latvian
Lithuanian
Mai Mai
Mano
Mende
Nuer
Oromo
Pashto
Persian
Polish
Rumanian
Russian
Serbian
Somali
Spanish
Temne
Teochew
Tigre
Tigrinya
Ukrainian
Vietnamese.
Access Translation & Interpretation Services
While most refugees learn English during their first few years in the United States, new arrivals and older refugees often need assistance in acquiring the knowledge and accessing the assistance needed for successful adjustment."
https://www.pa.gov/agencies/dhs/resources/refugees-in-pa/refugee-translation-interpretation
#metaglossia_mundus
"Mohamed Shueikh is an Egyptian translation supervisor and quality controller at the Cairo, Egypt-based Masreya Media.
In a WhatsApp interview on June 27, he shared his experience dubbing in Arabic the hit Netflix series "Squid Game."
The following are excerpts from the interview.
Briefly introduce yourself.
I am Mohamed Shueikh, an Egyptian translator with nine years of experience in translating and adapting dubbing scripts. I am the translation supervisor and quality controller at Masreya Media. For "Squid Game," my role was to direct and supervise translators and make strategic decisions on language type, character names and places.
How was working on the final season since "Squid Game" was Netflix's first K-drama to be dubbed in Arabic?
It was great news as l loved the series. I always wondered what it would look like if it were dubbed in Arabic.
Give a simplified explanation of the dubbing process.
It begins with receiving the Korean videos, script and their English-language translations. A detailed translation plan is developed, where we decide the style and tone of language. Then comes the translation of character names, places and games.
Afterward, the assigned translators begin their work and I later supervise them. After translation is complete, the text is sent to an Arabic linguist for review to ensure clarity and linguistic accuracy for the voice actors. The final stage is the quality control phase in which we check everything again.
What issues arose in translating the script from English instead of Korean?
The correct pronunciation of Korean names and terms requires thorough searching. Knowledge of the Korean language and culture was crucial for this process since translations in English are often written from a different perspective, frequently leading to misinterpretations.
It wasn't easy for us Arabic speakers to accurately express the sounds of Korean letters in writing so that our voice actors could deliver them correctly. But we did provide the cast with the correct pronunciations and contexts. I also remember the hard work needed to synchronize the dubbing voices and the actors' lips.
What did you discover about Korean culture after working on the series?
I find Korean culture rich and unique in language, tradition and arts. I'm fascinated by the invention of Hangeul (Korean alphabet), one of the world's simplest writing systems, and iconic dishes like kimchi and bibimbap (spicy rice mixed with vegetables and meat). What truly impressed me is the language's precise word usage and details like how the word for "sister" differs based on age and gender.
Jul 14, 2025
By Honorary Reporter Omnia Ameer from Egypt
msjeon22@korea.kr"
https://www.korea.net/NewsFocus/HonoraryReporters/view?articleId=274585&pageIndex=1
#metaglossia_mundus
"Of the 7,159 living languages in the world that the Ethnologue website now lists, it considers 3,193 to be “endangered,” that is, “It is no longer the norm that children learn and use this language.” That’s 45%. The Catalogue of Endangered Languages has compiled data on the “current vitality” of 3,394 languages, a slightly higher number. Ethnologue makes a further distinction: where “the child-bearing generation is no longer able to transmit the language to the next generation,” the language is “dying”; there are 1,030 of these, or 14%. It also reports that 454 languages have become extinct during the recent centuries of European expansion.
Wikipedia has an impressive (and depressing) list of extinct languages, starting with the most recent extinctions and running back several thousand years. It names three languages as having gone extinct in 2024, but since it cites an article in the Jakarta Globe that names eleven recent extinctions in Indonesia alone, it is surely not up to date. It tells us, however, that the last speaker of Columbia-Wenatchi in Northern Idaho died in 2023 at age 96; her name was Pauline Stensgar. The last speaker of Quapaw in Oklahoma died in 2022 at age 91; her name was Ardina Moore. The last speaker of Bering Aleut died in 2021 at age 93; her name was Vera Timoshenko. Some 95 languages, according to this list, went extinct in our century. Over 250 died in the twentieth.
Indepth Your Inbox
Get unbiased nonprofit watchdog news sent directly to your inbox.
Your Email
Before long there will be many more on the list, including the last remaining fluent speaker of Njuu, a “click” language of South Africa; her name is Ouma Katrina Esau, and she is now 92. Of 2183 living African languages, the Catalogue of Endangered Languages reports, 604 are considered endangered. About an eighth of the endangered ones are “critically” or “severely” endangered.
When these languages die, sometimes a similar dialect may live on, and may even flourish, but usually nothing remains except, with any luck, some tapes and transcriptions and studies by linguists, not usually enough to revive them. They are gone, and gone with them are their cultures: their songs, dances, ceremonies, games, jokes, stories, ways of seeing the natural world, and works of skill and art.
People speak misleadingly of another kind of extinction or death that is quite a different thing. Is Latin a dead language? I think no one speaks it today as a native or first language, though there are some who are fluent in it as a second language; the late Justice David Souter is said to have been quite good at it. So Classical Latin may be dead, or artificially sustained as “undead,” but it also lives on as the most widely spoken first language in the world, in the many dialects that we call Italian, French, Spanish, Portuguese, Romanian, Catalan, Sardinian, and dozens more. Old English is extinct, of course, as a spoken language, but its direct descendant is the indisputable world language today.
Languages constantly change, year by year. I don’t fully understand the variant of English my daughter speaks with her friends, but she is bilingual and can speak my language—my version of English—fluently. By the same token, if I were to meet Charles Dickens on the street today, I would understand him fairly well (I’ve read a lot of his books), but he would have great difficulty understanding me, unless I switched codes to nineteenth-century British. And so it goes, generation after generation, a perfectly normal evolution found in every tongue, though it goes faster in some languages than others. There’s no point in regretting it.
But it is a mournful fact that every month or so the world loses a language completely, a language with no offspring.
It is heartening to learn, however, that in many countries there are programs dedicated to teaching children to speak their elders’ dying language. To mention a few efforts in the US: There are some eighty fluent Chippewa Cree speakers in Montana, and some of them are active in classes and activities for children. The University of New Mexico has a project where linguists and other scholars are working with native speakers of Jicarilla Apache. The Wampanoag Language Reclamation Project had to deal with fact that their were no elders who spoke Wampanoag (or Massachusett), so Jessie Little Doe Baird, who co-founded the Project in 1993, got a degree in linguistics from MIT in order to help her reconstruct it. There were opaque transcriptions by a white missionary, but not much else to go on, so she studied a few related Algonquian languages still surviving in New England and worked out what must have been the lost words and structures of Wampanoag. After thirty years the Project has born fruit, for there are now some fluent young native speakers.
Projects like these cost money, and local tribes seldom can afford them alone. The Chippewa Cree program costs about a million dollars a year, nearly all of it from Federal sources. In December 2024 the Biden administration announced a ten-year National Plan on Native Language Revitalization, calling for a $16.7 billion investment. Donald Trump, who wants to make English the only official language of the United States (though he can barely speak it himself), is not likely to support this plan, any more than he supports efforts to rescue endangered species. The Federal Government may well relapse to its vicious policy of “killing the Indian to save the man.” As for Trump, he would no doubt like to deport all the Native American tribes, with their languages, back to the countries he thinks they came from.
I am happy to hear from readers with questions or comments: mferber@unh.edu.
Michael Ferber moved to New Hampshire in 1987 to join the English Department at UNH, from which he is now retired. Before that he earned his BA in Ancient Greek at Swarthmore College and his doctorate in English at Harvard, taught at Yale, and served on the staff of the Coalition for a New Foreign Policy in Washington, DC. In 1968 he stood trial in Federal Court in Boston for conspiracy to violate the draft law, with the pediatrician Benjamin Spock and three other men. He has published many books and articles on literature, and has a deep interest in linguistics. He is married to Susan Arnold; they have a daughter in San Francisco."
Michael Ferber
July 12, 2025
https://indepthnh.org/2025/07/12/speaking-of-words-language-death/
#metaglossia_mundus
"Voice Coaching for Interpreters
Saturday, November 22, 2025
Organizer:
Location: Rue des Savoises 15 La Maison Internationale des Associations Geneva 1205 CH
Workshop Description
One of the most crucial tools for conference interpreters is their voice. As interpreters, our voices are the primary means of conveying messages and communicating effectively. A well-trained voice enhances clarity, projection, and endurance—essential qualities for simultaneous interpreting. The better our voices sound, the more pleasant and professional the experience will be for our clients.
However, the voice is a delicate instrument, requiring proper care and technique to withstand long working hours under extreme stress—especially in the current landscape where remote simultaneous interpreting is becoming increasingly prevalent. Despite its importance, voice training is often overlooked in interpreter education, leaving many professionals without the essential knowledge to maintain and optimize their vocal performance.
Workshop Structure
This workshop is designed in two parts:
1. Group Training:
Participants will receive an introductory session on fundamental vocal techniques, including:
Proper breathing methods, utilizing the diaphragm to create a well-supported airflow for effortless vocal production.
Breathing and articulation exercises to strengthen diaphragmatic support.
Posture alignment in the interpreting booth to open up resonators and enhance voice quality.
Correct placement of consonants and vowels across different languages.
Voice projection techniques to ensure interpreters can speak at an optimal pitch and volume without strain—helping to prevent vocal fatigue during long workdays.
2. Individual Coaching:
Each participant will receive personalized guidance, including:
A voice diagnosis to identify any existing issues or inefficient habits.
Tailored corrections to improve voice placement, breathing, and projection.
Real-time demonstrations where fellow participants can observe and learn from individual coaching sessions.
Registration fee:
Registration fee for members of AIIC - CHF 100.00 (EURO 108.00)
Registration fee for non-members of AIIC - CHF 200.00 (EURO 216.00)
https://aiic.org/client/event/roster/eventRosterDetails.html?productId=756&eventRosterId=45
#metaglossia_mundus
Judge Dana Sabraw this week ordered officials at Otay Mesa Detention Center to provide him with a Mongolian Sign Language interpreter.
"by Wendy Fry • CalMatters
July 13, 2025, 1:38 p.m.
A deaf Mongolian man has spent more than four months in a Southern California immigrant detention center without the opportunity to communicate with anyone who understands Mongolian Sign Language, according to his civil rights attorney.
“He’s basically been in solitary confinement because he has not had one person actually speak to him in Mongolian Sign Language for the entirety of the time that he’s been in proceedings and detained,” said his attorney, Alegría De La Cruz, director of litigation for the Disability Rights Legal Center.
U.S. Southern District of California Judge Dana Sabraw this week ordered officials at the Otay Mesa Detention Center to provide him with a Mongolian Sign Language interpreter.
The judge also directed immigration authorities to redo two assessments that could affect his request for asylum. One would examine his mental health, and the other would evaluate whether he has a credible fear for his safety if he returns to his country.
“How can he meaningfully participate if he doesn’t know what’s being said and he cannot communicate?” Sabraw asked a federal attorney at the hearing on Wednesday.
Assistant U.S. Attorney Erin Dimbleby said many people don’t fully understand the legal proceedings in immigration court.
The man’s family requested that CalMatters identify him by the name Avirmed because of their fear that he could be harmed by the Mongolian government if he is returned to his home country.
Avirmed’s detention after seeking asylum underscores the sharp shift in border policies from the Biden administration to Trump’s.
Under Biden, asylum seekers who were not threats to public safety were often released on bond rather than being detained while their cases moved through immigration court. The Trump administration has taken a much stricter approach with detention numbers reaching record highs – a tactic his supporters say is working. Since President Donald Trump took office, unauthorized border crossings have plummeted to historic lows.
But Avirmed’s representation by the Disability Rights Legal Center also shows how some immigrant advocates are changing tactics in response to the Trump administration’s immigration crackdown. In this case, they’re drawing on federal disability laws that prohibit discrimination against people with disabilities by any federal program, including the immigration court system.
The Trump administration recognizes its “own power, and it can be very dangerous unless someone checks it the way the judge did,” said Sylvia Torres-Guillen, the president and CEO of the organization
Key ICE interviews without interpreters
Avirmed left Mongolia early this year and entered the U.S.in February seeking asylum from persecution because of his disability. A 2020 assault in Mongolia left him with a traumatic brain injury that causes seizures and memory loss. He was attacked because of his disability, according to court records. His family declined to say how he reached the U.S.
A legal complaint filed on his behalf says Avirmed gave border officials a letter written in Mongolian and translated into English, notifying them of his disability and his intent to seek asylum. Customs and Border Protection agents refused to read or accept the letter, his attorneys allege in the complaint against the Department of Homeland Security.
Agents transferred him to Immigration and Customs Enforcement, where he was placed in detention at the for-profit Otay Mesa Detention Center, run by CoreCivic, where he is still being held.
Agents interviewed him without an attorney or sign language interpreter and tried using Google Translate to ask him if he feared returning to Mongolia, according to the complaint. They badly misunderstood him, identifying his sponsor as a daughter named Virginia Washington when he does not have a daughter, according to the lawsuit. His sponsor is his sister, who lives in Virginia.
Avirmed also underwent a mental health evaluation without any interpretation, which the judge ordered ICE to repeat, saying he “has a right to be involved where he understands and can respond and communicate, and be part of the process, not a bystander.”
California officials have been critical of mental health resources in ICE detention centers. By law, they have access to inspect federal immigration detention facilities. An April report from California’s Attorney General’s office documented what it described as severely inadequate mental health care services inside ICE facilities.
“No facilities consistently offered adequate psychotherapy services for the mental health conditions most commonly observed in detainee populations in California,” the report states, naming detention centers in Imperial and San Diego counties that California officials found to have behavioral health staffing vacancies.
The companies that operate the detention centers contested the attorney general’s findings, with one calling the report an example of a “politicized campaign” to interfere with deportation efforts.
‘Like Greek to me’
Avirmed has no criminal record. According to ICE’s data, which shows less than 10% or 125 out of 1,350 people currently detained at Otay Mesa have been convicted of a felony or misdemeanor.
Sabraw, the judge, agreed with De La Cruz and Avirmed’s other attorneys that the 48-year-old man was likely unable to understand what was happening during earlier proceedings, so they should be repeated with an interpreter and in a language he understands.
“He has a right, doesn’t he? To be able to fully participate in any significant proceeding?” Sabraw asked Dimbleby.
“MSL (Mongolian Sign Language) is not a super common language,” Dimbleby argued at one point.
For future proceedings, the federal government had proposed providing Avirmed a “relay team” that would allow him to testify through a certified deaf interpreter. The interpreter would then translate the testimony into American Sign Language, and then the ASL would be translated into spoken English.
“Under this relay proposal, everyone would understand except for Mr. Avirmed,” Sabraw pointed out.
The judge said the government’s plan does not allow Avirmed to participate in court proceedings because he does not understand English or ASL.
“That’s like speaking Greek to me,” Sabraw said.
CalMatters is a nonpartisan and nonprofit news organization bringing Californians stories that probe, explain and explore solutions to quality of life issues while holding our leaders accountable."
crime/2025/07/13/deaf-mongolian-immigrant-held-by-ice-in-otay-mesa-for-months-without-access-to-interpreter/
#metaglossia_mundus
The 56th Modern Korean Literature Translation Awards is now open for entries. Translators from around the world are invited to submit English editions of contemporary Korean literature before the end of August.
"...The 56th Modern Korean Literature Translation Awards is now open for entries. Translators from around the world are invited to submit English editions of contemporary Korean literature before the end of August.
Since 1970, The Korea Times has promoted the global reach of Korean literature. It seeks not only to translate acclaimed literary works but also to nurture aspiring literary translators. The annual competition calls for submissions in two categories -- fiction/drama and poetry.
Supported by the Ministry of Culture, Sports and Tourism and KB Financial Group, the Modern Korean Literature Translation Awards categories are:
Grand Prize in Fiction/Drama: 7 million won (around 166,000 baht)
Grand Prize in Poetry: 4 million won
Commendation Awards: 2 million won for both categories
Kevin O'Rourke Award: 1 million won to an entry from either category
Applicants may send a translation of 1) a work of fiction (a novel, novella or short story) or a play, or 2) 10 poems by the same writer. Applicants should submit only one entry in either category. Translations should be sent before Aug 31.
Last year, Wingshun Pang won the Grand Prize in Fiction for translation of Kim Cho-yeop's short story Why Don't The Pilgrims Come Back, while Julie Sohn received the Grand Prize for the translation of Choi Seung-ho's poetry collection The Snowman Suicide Incident.
Korea is home to a plethora of literary works. The Modern Korean Literature Translation Awards reflects Korea's push for internationalisation across diverse fields. With Han Kang's Nobel Prize in Literature 2024, Korean literature has achieved the highest international acclaim. The Vegetarian was her first novel to be translated into English. It received the International Booker Prize in 2016, which helped expand Han's readership worldwide.
For more details, visit koreatimes.co.kr.:
THANA BOONLERT
Please credit and share this article with others using this link: https://www.bangkokpost.com/life/social-and-lifestyle/3067162/korean-literature-translation-awards-seeks-entries. View our policies at http://goo.gl/9HgTd and http://goo.gl/ou6Ip. © Bangkok Post PCL. All rights reserved.
https://www.bangkokpost.com/life/social-and-lifestyle/3067162/korean-literature-translation-awards-seeks-entries
#metaglossia_mundus
"“Barron’s China” has been relaunched this month, seeking to help global investors understand the complex Chinese wealth and investment ecosystem.
By PA/TPN, in Asia, World, Economy
This arrangement shifts a 2018 translation deal with the business outlet Caixin, which also ensures that Chinese-speaking investors can get the best and most reliable information in their native language.
In collaboration with TMTPost, Dow Jones will be producing the publication, promising to be a bridge connecting China and the world, a fulcrum for the globalisation of Chinese
https://www.theportugalnews.com/news/2025-07-12/translation-deal/99215
#metaglossia_mundus
"In defence of translation
Given the obvious power difference between Welsh and English, and our long colonial relationship, is it right or wrong to translate English works into Welsh?
We’ve been extremely pleased with the overall reception we have had since setting up Melin Bapur at the end of 2023 and launching in February 2024.
Dozens of our customers have gone out of their way to give us great feedback and the support we’ve had from Welsh institutions like Radio Cymru and nation.cymru has been great; even better has been the way talented authors and translators like Anna Gruffydd, Mary Burdett-Jones, Ian Parri, Peredur Glyn, Sharon Morgan and Richard Crowe have approached us and paid us the enormous compliment of entrusting us with publishing their work.
Nevertheless I have been fascinated to note the occasional more negative comment we have received as well – nothing genuinely nasty or insulting, thankfully, but some comments specifically arising from our decision to publish Welsh translations of English books like The Hobbit, The Time Machine, The Call of Cthulthu, The Vagina Monologues and the like – these comments have questioned either the value or the appropriateness of such things, and particularly, accusing us of somehow undermining the publication of original literature in Welsh.
It’s a fair question to ask. Given the obvious power difference between Welsh and English, and our long colonial relationship, is it right or wrong to translate English works into Welsh?
Arguments against
Based on the comments we’ve received it seems that the arguments against translation from English, broadly speaking, are that:
1) people will buy and read translations instead of original books by Welsh writers, thereby harming Welsh authors;
2) there’s no point as Welsh speakers generally can read English anyway, so they won’t want to read translations.
Keen observers will note that these two arguments are contradictory: if nobody wants to read them then how can they harm Welsh writers?
Regardless, it’s pretty easy to disprove the second point with the fact that people do buy and read these books. This argument is based on a misunderstanding of why people buy and read books (in Welsh or otherwise), which probably arises from a prejudice against translation as a process and an assumption that reading a text that has been translated is an experience that is on some essential level inferior.
Holes
This is not, however, how most people actually think. Needless to say we don’t know everything about all our customers, but I would be willing to bet that the majority have already read these books in English, especially Yr Hobyd; they want them in Welsh not because they think that a Welsh version will be better (or worse), but because it will provide a different sort of experience. If anything the fact they already know the book is the whole point.
But we’ll come back to that, after addressing the idea that these books threaten or harm original writing in Welsh.
The idea seems to be that readers will choose to buy and read Tolkien or Lovecraft or Eve Ensler instead of Welsh authors; therefore we are hurting them by publishing these books.
The translated works and H. P. Lovecraft
This argument might make sense on first glance, but the more you think about it, the more holes appear.
The argument seems to take it for granted that the demand for books in Welsh is some kind of fixed quantity: that buying and reading any one book in Welsh means one doesn’t buy or read another book in Welsh. Publishing, under this assumption, is a zero-sum business.
This might be true of some products, like, say, washing powder: if I buy one brand I don’t need to buy another.
But reading isn’t like that. Even it were, this argument betrays a lack of confidence in the value and appeal of the same original Welsh books it purports to defend.
Do we really think that the only reason people read original Welsh books is that they can’t get translations of the books they really want to read? And that – in order to protect original writing in Welsh – we need to restrict readers’ access to other books in Welsh, in case they might prefer them?
I believe that there is more value to Welsh literature than that!
Opening doors
Of course in reality, Welsh books already have to compete with Tolkien, J. K. Rowling and the rest, regardless of whether or not they’re available in Welsh; just as they compete with the television and social media and Netflix and the gym and everything else people choose to do in their spare time.
In fact Welsh readers already read Tolkien, but before we brought out Yr Hobyd they were doing it in English.
Sure, bring out a new book in Welsh (whatever it is) and it might mean someone doesn’t read another book in Welsh. But it might also mean they read one fewer book in English, or that they spend less time doing something else, and more time reading in Cymraeg. The book Yr Hobyd is probably most likely to push off someone’s reading list is The Hobbit!
(I’ve been accused of Thatcherism for making this argument!)
And this brings us to the point of publishing these books, which is the very real possibility that they are actually bringing people into reading in Welsh who wouldn’t otherwise do it.
Many of our customers have contacted us to say exactly this; many of them are learning Welsh and want the book as an exercise to improve their Welsh (which of course is another great reason why it’s important we have these books).
Younger readers, too, or other fluent speakers who don’t currently read at all in Welsh, but want to, yet don’t feel the current offer in Cymraeg caters for them, or don’t know where to start, or worry that they’ll struggle to understand and want the crutch of a familiar book.
Gateway
Of course there’s no guarantee these people will go on to read another book in Welsh, but even if they don’t, we’ve helped them to read a book in Welsh they would not otherwise have read, which is surely valuable in itself.
If this is just the start, then better still. The other half of our mission with Melin Bapur is to republish the (original) Welsh literature of the past, and to use the translations as a ‘gateway’ into this exciting world.
Perhaps I haven’t convinced you, but that’s ok: you don’t have to buy our translations (no doubt that’s me being Thatcherite again?), but do check out the original Welsh books that represent about three quarters of Melin Bapur’s offer!
It’s possible to overstate the argument, and I don’t wish to be misunderstood: something extremely valuable would be lost if the only Welsh books available were translations.
I want to see original Welsh writers supported too, and this is one of the reasons we don’t compete with the traditional Welsh publishing industry for Books Council grants (which don’t seem to be offered for translations anyway).
But the current situation, where virtually nothing is published in translation for adults, is so far away from this nightmare scenario that it is if anything too far in the opposite direction.
Publishing popular stories are what normal, healthy, living languages do, and our reluctance to do so in Welsh seems to speak more of our own insecurities than anything else.
There is more than enough room in Wales for the occasional Hobyd – after all, they’re famously quite small!"
Adam Pearce, Editor, Melin Babur Books
12 Jul 2025
https://nation.cymru/feature/in-defence-of-translation/
#metaglossia_mundus
"China publishes Han-Tibetan version of major dictionary
XinhuaJuly 12, 2025
China attaches great importance to the use of the Tibetan language, as evidenced by the recent publication of a key bilingual dictionary.
On Friday, the Han-Tibetan version of the Modern Chinese Dictionary, compiled and translated over 13 years, was jointly published by China Tibetology Publishing House and The Commercial Press, according to the ethnic and religious affairs commission of Qinghai Province.
Against the backdrop of rapid social development, Tibetan people have shown a growing desire to learn the national standard Chinese (Han language) while also preserving their own. However, authoritative and user-friendly Han-Tibetan reference books have been scarce, prompting Qinghai -- a province with a large Tibetan population -- to launch the translation project.
The three-volume dictionary has over 70,000 Chinese entries and 200,000 Tibetan entries, totaling more than 10 million characters. While featuring vocabulary, explanations and example sentences with concise, accurate and practical translations, it also integrates word forms and meaning transformations in line with Tibetan grammatical rules.
This dictionary serves as an authoritative language tool in Tibetan-inhabited areas, providing valuable resources for education and cultural exchange and contributing to China's efforts in promoting ethnic exchanges and integration, according to the commission.
http://www.china.org.cn/2025-07/12/content_117975656.shtml
#metaglossia_mundus
"...Nashwa Nasreldin on why Arabic literature translators are needed now more than ever
Editor and author translated Sheikh Zayed Book Award winner Abdelrashid Mahmoudi's After Coffee, among other works
Nilosree B
July 11, 2025
Nashwa Nasreldin believes the work of translators in literature is now more urgent than ever.
"The opportunity now lies in uplifting the work and voices of translators based in the Middle East, who have less access to the support networks than we do living in the West," says the translator, editor and author with a mission.
Born in Kuwait to Egyptian parents, Nasreldin, who currently lives in Suffolk in the UK, continues: “Recently, we held our inaugural workshop for emerging translators in Gaza, run by the ArabLit platform, with each session led by experienced professional literary translators - some of the best in the field of Arabic/English literary translation.
"We need more translations to drown out the voices of those who try to rewrite our stories," she tells The National. "There has been an outpouring of literature from and about Gaza and Palestinians more broadly since October 7, 2023. This shows a recognition in the important role literature plays in recording, reporting, archiving and legacy-making."
Nasreldin's translated many works into English, including After Coffee by Egyptian author and academic Abdelrashid Mahmoudi, who won the Sheikh Zayed Book Award in 2014.
Being tasked to translate Mahmoudi's book was a big moment for her as her first solo book project, she says.
“I really enjoyed Mahmoudi’s book when I read it in Arabic. For me, literary quality is an important factor when I come to choose a project given that you have to inhabit the book's world so intensely in the process of translating it."
Her work has since take her around the world, including to the Jaipur Literature Festival in February – where she spoke about her struggles and the importance of translators.
Nasreldin's career as a translator began when she studied Moroccan writer Mohammed Bennis’s work as a part of her master's course work at the Vermont College of Fine Arts in the US.
“I studied Bennis and produced an in-depth study of his writings on translation. At this time, I also met writer and translation-advocate Maureen Freely, who was a guest lecturer, and who introduced me to the British Centre for Literary Translation and their emerging translator mentorship. When I moved to the UK in 2013, I applied and was awarded the mentorship, which then propelled my career.”
Nasreldin is currently translating two short books that she says she's very passionate about.
A Brawl in Jahannam is a novella by award-winning Libyan author Mohammed Alnaas, whose debut novel, Bread on Uncle Milad's Table, won the 2022 International Prize for Arabic Fiction. It’s a satiric retelling of events that the narrator claims took place in the 1990s in a small Libyan village named Jahannam, which literally translates to hell.
The other book is Side Entrance to the House by Omani writer Amal Alsaeedi, in which the author’s ancestral house serves as a vehicle to unearth memories and images of childhood and youth, triggering philosophical reflections on her troubled relationships with her husband, parents and siblings.
"These books represent a new generation taking bold literary risks, reinterpreting classical Arabic styles," says Nasreldin.
But making a living as a translator is not easy, and many in the field are forced to find several sources of income, she adds.
"I split my time working as a writer, an editor and a translator," she says. "Like most emerging translators, initially the challenges lay in making a name for oneself in a very competitive industry – and this was back when there were fewer Arabic literary translators than there are now. At the time, it was also difficult to compete with translators who didn't have an Arab-sounding name, as publishers would often assume we were less proficient in English."
That became less of an issue in recent years, thanks to a demand for diversity that has helped tip the scales more towards equity, she adds.
"But the disparity still exists, especially amongst Arab publishers and authors who still believe that a so-called 'native English speaker' would have a stronger grasp in English than a translator who was raised bilingual, as I was.
"In the field of Arabic literature in translation, we only have a handful of prizes including the renowned Saif Ghobash Banipal Prize for Arabic Literary Translation and the newer Bait AlGhasham DarArab Translation Prize. Also, the ArabLit Short Story Prize is particularly impactful as it supports writers who may not have had the opportunity to publish book-length work.”
But the publishing industry as a whole is changing, albeit slowly.
"The work of translators are increasingly recognised, thanks to social media campaigns like the #namethetranslator, which seeks to ensure that translators are named alongside their work," Nasreldin says.
"There are dozens of excellent Arabic translators working in the field today. A group of us communicate regularly on mailing lists, where we discuss translation quandaries, share opportunities, celebrate successes and band together when there is a need to address a threat to our rights as creators – especially with the unregulated rise of mechanical translation, having a supportive community matters.
"I think, we will see a rise in translations of Arabic literature from the Middle East and North Africa region as more translators are trained and gain confidence in an industry that had been relatively opaque previously," she adds."
https://www.thenationalnews.com/arts-culture/books/2025/07/11/nashwa-nasreldin-on-why-arabic-literature-translators-are-needed-now-more-than-ever/
#metaglossia_mundus
"EC suspected own translator to be alleged Russian spy, report states
Carl Deconinck
A translator who was present at closed-door top meetings with Ukrainian President Volodymyr Zelensky in late 2024 was suspected by the European Commission of being an alleged Russian spy.
She was caught taking notes, French newspaper Le Monde reported on July 11, which was strictly forbidden in such confidential high-level meetings.
The incident reportedly took place on December 19, 2024, during a European Council meeting in Brussels, with Ukrainian President Volodymyr Zelensky in attendance.
It happened one month after the election of US President Donald Trump, who had campaigned on ending the Ukrainian war quickly. European leaders were looking for a shared stance on Ukraine to ensure any potential peace deal would not be detrimental to the country.
Belgian authorities were informed of the findings of the internal investigation the EC had and they would now have to decide what course of action to take, if any, regarding the case.
It was Czech interpreters who reportedly noted the allegedly suspect translator taking notes.
They reported it to security services, who caught her doing so. They seized all her materials and escorted her off the premises, Le Monde reported.
On the same day, an investigation was opened, while her accreditation was revoked and she was no longer allowed to enter EC buildings.
The translator, described as “Ms I” by the French newspaper, was reportedly born to Russian parents. She worked as interpreter, for 20 years as a freelancer for NATO, the EC and the French ministries of defence and foreign affairs.
“Ms I trained at the Institute of Translators, Interpreters and International Relations in Strasbourg after earning degrees in English, foreign literature and psychology in Kyiv in 2002,” Le Monde stated.
In a reaction to its report, the EC admitted: “An incident involving note-taking, which is prohibited by our code of conduct, did indeed occur during the December 19, 2024 meeting.
“Ukrainian interpretation was necessary for this meeting because of President Zelensky’s participation. The notes were confiscated. After a careful review of the facts, it was decided to no longer use the interpreter’s services,” Le Monde wrote.
While her work for the French Government was limited, when contacted by Le Monde, French authorities reportedly said they would “draw all the necessary consequences from this incident”.
The General Directorate of Internal Security (DGSI), though, responsible in particular for counter-espionage, said it had not received any information about the case.
Despite the December incident, the woman in question remained listed as an accredited interpreter for NATO, the French Permanent Representation to the European Union and the French ministries, according to Le Monde.
Ukrainian embassies in France and Brussels reportedly did not allow her to be present during visits of Zelensky for some time, claiming she allegedly maintained professional relationships with Russian authorities.
In a reaction to Le Monde, the translator said she was “very surprised to be approached about a matter of no interest”. She argued that her continued work with NATO and French ministries proved this was simply a misunderstanding.
She refused to comment on the note-taking incident, stating: “Everything to do with my work as an interpreter is covered by confidentiality and my entire professional life has been marked by respect for professional ethics alone.”"
https://brusselssignal.eu/2025/07/ec-suspected-own-translator-to-be-alleged-russian-spy-report-states/
#metaglossia_mundus
"...Interpreters for Morgan County Fair concerts will be on a raised platform on the west side of the stage, to the right when viewing from the crowd.
Courtesy of the Morgan County Fair
The Morgan County Fair is introducing ASL interpreters at its concerts, with the goal of expanding access for deaf and hard-of-hearing concert-goers.
Study Suggests Lyrics Have Gotten Simpler Over Time
Two ASL licensed interpreters will alternate throughout the fair's concerts today and Saturday, providing interpreter services live during the concerts.
The goal is to provide better access to the fair and fair entertainment for everyone in the community.
Fair Board President Gary Hadden said this is something they have wanted to do but were not sure where to start with the process. This year, they were able to work with the artists' managers and booking agencies, as well as members of the Deaf community, to make a plan to provide interpreting services.
"We have had requests for an interpreter and so we asked our artist managers and booking agencies to see what we need to do and what the requirements were to meet their approval and provide the service," Hadden said.
The interpreters will be on a raised platform on the west side of the stage, to the right when viewing from the crowd. They will be visible from both the Big Ticket area and the general admission area, so those needing the service do not have to purchase one of the higher-cost tickets.
When the call was put out for interpreters, there are many who volunteered, Hadden said.
"As a not-for-profit organization, we received a lot of volunteers," Hadden said. "We sought out people willing to volunteer, some were licensed and some were not."
To meet requirements, the interpreters have to be licensed.
Kate Van Valey, an office assistant for the fair and a teacher at Illinois School for the Deaf, helped develop the plan and connect the fair to resources available in the community.
"We are on a journey to meet the needs of the community," Van Valey said. "It's a growth process."
Support for the process has been positive, Van Valey said.
"We had members of the Deaf community share their ideas and their response has been really positive and they have been encouraging people to come, knowing that they'll have access," she said.
The interpreters providing services have experience interpreting music, which is different from interpreting standard conversation, Van Valey said.
"There is a shortage of interpreters, especially ones who are comfortable performing in front of a large crowd with music," Van Valey said. Those interpreting music use their entire bodies and facial expressions to help translate the music and lyrics.
The fair board worked with the entertainment agencies to provide resources to the interpreters to help improve the experience for those who are deaf or hard of hearing.
Several measures are being put into place, including a set list and a direct feed into the concert to help avoid outside noise that could cause problems for the interpreters.
"If the performer is ad-libbing, the interpreter can hear that better and provide that," Van Valey said. "With the direct vocals in their ear, they have better access to that portion and not just the songs. Their engagement is better accessed."
The effort is something they hope to expand, improve and continue into future years, Hadden said.
"This is our first year, so we are asking people to be patient as we are learning as well," he said.
July 11, 2025
Samantha McDaniel-Ogletree
REPORTER
https://www.myjournalcourier.com/news/article/morgan-county-fair-asl-interpreters-20764456.php
#metaglossia_mundus
Did you know? The Translating Division of the Office of Language Services provides translation services in some 140 language combinations
"The Translating Division of the Office of Language Services provides translation services to the Department of State, the White House, and other U.S. Government agencies. We assist in handling the foreign-language components of the written record of diplomacy: correspondence, treaties, reports, speeches, course materials, briefing slides, biographical sketches, conference agenda, media items, laws, and forms. The team of staff translators, assisted by a corps of vetted contractors, offer their services in some 140 language combinations. LS translators work closely with negotiators when certifying that foreign language versions of treaties and international agreements have the same meaning as the English—a painstaking process that requires attention to nuance and the ability to separate linguistic issues from policy differences. Typically, several rounds of certification are needed to achieve substantive conformity.
A translation project may also be a website, a set of subtitles, a desk-top published brochure, an embedded .pdf image, a handwritten sworn statement, or an audio transcript. Helping the Translating Division cope with the demands of these formats is a team of Translation Project Managers who coordinate each assignment, from initial intake to final delivery, and who will help with all your logistical concerns. U.S. Government agencies can request assistance with translating projects by emailing the Translating Division.
Are you interested in working for us? Please visit our Information for Freelance Linguists page. In the Translating section, you will find information about translating for the Office of Language Services, as well as translation test guidance. Permanent direct-hire interpreter and translator positions are very rare. When available, openings for staff positions will be published on the USAJobs website.
To request translation (written) assistance for documents and agreements, please e-mail the Translating Division."
https://www.state.gov/about-us-ols/translating
#metaglossia_mundus
"...Run For The Bibleless flag off
The Church in Kenya has been called upon to support the translation of the Bible into lesser-known languages, a project spearheaded by Bible Translation and Literacy (BTL).
Speaking during the 5th edition of the “Run for The Bibleless” event held at Mama Ngina Waterfront in Mombasa, BTL’s Board Chair, Kendi Ogamba, highlighted the ongoing need for communities to access the scriptures in their native languages.
Ogamba emphasized the high costs associated with Bible translation and urged the Church to contribute to this cause through the charity walk, “Run For The Bibleless.”
“We shall continue with the Run For The Bible until all communities receive the scriptures in their own language.” Ogamba said.
Founded in 1981, BTL is a Christian-based organization dedicated to Bible translation and sustainable literacy development programs for small language groups in Kenya and beyond.
To date, the organization has completed and dedicated six full Bibles and twelve New Testament translations. Currently, BTL is working on eleven translation projects across the country, including those for the Ribe, Jibana, Kambe, Kauma, and Rabai languages, primarily in the coastal region.
The “Run for the Bibleless” initiative, started in 2007 in Nairobi, expanded to Mombasa in 2021. The event now takes place in five major cities in Kenya: Nairobi, Mombasa, Nakuru, Eldoret, and Machakos.
BTL’s Board Chair, Kendi Ogamba (right), President of the Court of Appeal (second right), and others.
The President of the Court of Appeal, Daniel Musinga, noted that the funding gap left by missionaries should no longer be felt.
“It’s important for every community in the country to receive the Bible in their language. Kenyans can play a vital role in funding this project.” Musinga said.
Musinga further remarked that translating the Bible into local languages would assist the judiciary in fulfilling its mandate, as many rural residents struggle to communicate in either English or Kiswahili.
“These translations will help eliminate the language barriers experienced in courts,” he explained.
He also encouraged parents to allow their children to learn and use their mother tongues, expressing concern over instances where children, like Mutiso, do not speak their native Kamba language, despite evidence that children under the age of ten years can learn multiple languages.
BTL provides various literacy programs focused on developing language materials in mother tongues and training teachers and supervisors.
Currently, BTL’s literacy initiatives are undertaken within the Duruma, Digo, Pokomo, Orma, and Waata languages.
The organization has received recognition from UNESCO and the Directorate of Adult and Continuing Education for its significant contributions to literacy development"
Raymond Zaka
July 13, 2025
https://www.kbc.co.ke/church-in-kenya-urged-to-support-bible-translation-efforts/
#metaglossia_mundus
"By Janette Toral
July 13, 2025
EDUCATION is undergoing a significant transformation with the rise of artificial intelligence. Classrooms are beginning to test and use AI-powered tools. Students are turning to chatbots not just for brainstorming, but also for writing, summarizing, and analyzing. This raises a vital question for educators: When AI can generate content and assess performance, what remains uniquely human in the learning process?
I recently had the opportunity to speak at MoodleMoot Philippines 2025, where digital learning advocates gathered to explore how AI is transforming education. I emphasized the importance of keeping the human in the loop, especially when we use technology. While AI can assist and accelerate the learning process, it is human judgment, context, empathy, and ethics that must ultimately guide it.
AI can help personalize learning, offer real-time feedback, provide language translation, support learners with disabilities, and automate content generation. These benefits, however, come with challenges that educators can no longer overlook.
Many AI experts warn that certain entry-level jobs may soon become scarce. Tasks such as research, transcription, scheduling, basic content writing, and social media management are increasingly being handled by AI systems. These tasks once served as stepping stones for young professionals to build critical thinking, effective communication, and decision-making skills.
The question now becomes: How can schools prepare students for a future where these traditional starting points may no longer exist?
Rather than shielding learners from AI, education must empower them to rise beyond its limitations. The focus should shift toward developing distinctly human skills that machines cannot replicate. These include the ability to solve problems collaboratively, reason ethically across cultures, communicate persuasively, build strong one-to-one and group relationships, and lead with vision and compassion. These traits will help define the human edge in an AI-saturated environment.
Keeping the human in the loop also means guiding both educators and learners in the responsible and intentional use of AI. The goal is not to replace teachers but to enrich the learning experience. Educators play a key role in reviewing and contextualizing AI-generated content, teaching students to ask meaningful questions, and verifying the credibility of AI outputs. They can also integrate AI literacy into subjects such as digital citizenship or media ethics.
For students, AI should serve as a starting point, not a final answer. By practicing how to rewrite chatbot-generated content for clarity, credibility, or empathy, students strengthen their communication skills. They also gain critical insight when comparing responses from different tools and identifying potential biases or errors.
Some schools are taking a proactive approach by developing their custom AI chatbots trained using local curriculum, school values, and preferred instructional styles. These models, similar to tools like Khanmigo, are designed to encourage questioning rather than simply delivering answers. This promotes deeper understanding while giving schools the ability to monitor usage, refine content, and align technology with educational objectives.
To truly prepare students for success, institutions must go beyond digital literacy and begin fostering AI resilience. This means helping learners adapt, reskill, and thrive in a world where automation is rapidly evolving. Curriculum development should emphasize project-based learning, real-world simulations, mentoring, creative entrepreneurship, and scenario-based training using AI tools. These experiences help students demonstrate originality, take accountability, and exercise leadership, all of which remain irreplaceable by machines.
As I emphasize in my book “Smarter with AI,” we must strike a balance. The future is not a choice between humans and machines. It is about using AI to elevate our human strengths: creativity, compassion, and critical thinking.
We must not only teach students how to use AI; we must teach them how to lead with it."
https://www.manilatimes.net/2025/07/13/business/sunday-business-it/rethinking-learning-as-ai-redefines-the-first-job/2148233
#metaglossia_mundus
"Inch Press launches new zine to bring Armenian literature to English-speaking readers
International Armenian Literary AllianceJuly 12, 2025 Last Updated: July 12, 2025
Fiction writer, essayist and translator Garen Torikian
Inch, a new quarterly literary zine, launches with a mission to bring Western Armenian literature to English-speaking readers through high-quality literary translation. Each issue is devoted to a single story—carefully chosen and thoughtfully translated—allowing readers to focus on its full emotional and cultural range.
Founded by writer and translator Garen Torikian, Inch is dedicated to presenting Armenian stories in all their multifaceted depth: satirical, surreal, lyrical and everything in between. “So much Armenian literature exists beyond the historical or tragic lens through which it’s often filtered. I started Inch to share stories that are sharp, funny, strange, intimate—stories that reveal how richly varied Armenian writing is.”The debut issue features a sharp, satirical piece by Yervant Odian, a tale of delusion and trickery, revolving around the creation of a new dictionary. True to Inch’s mission, the translation doesn’t merely replicate the story across languages: it creates a space where translators are visible and valued as literary artists, bringing their interpretive choices and cultural insight to the forefront.
“While I have been moved while reading and translating other works of Armenian literature, I cannot say I have often been entertained in the most basic sense of the word,” says translator Jennifer Manoukian. “Armenian literature does not always need to be heavy or didactic. It can also bring some laughter and lightness to a reader’s day. Odian’s work does just that.”
Supported by a grant from the Calouste Gulbenkian Foundation, Inch publishes both celebrated and emerging voices, with the aim of fostering connection within the Armenian diaspora and beyond.
Garen Torikian is a fiction writer, essayist and translator from Western Armenian into English. His writing has appeared in Wasafiri, Guernica, Electric Literature, Nexus, Potomac Review and SPEAK. In 2021, he edited Mo(a)t, an anthology of contemporary Arabic stories in translation. He has an MFA in Fiction from Columbia University in April 2022, and an MA in Literary Translation from the University of East Anglia."
https://armenianweekly.com/2025/07/12/inch-press-launches-new-zine-to-bring-armenian-literature-to-english-speaking-readers/
#metaglossia_mundus
Large language models don’t “think” like humans—but their internal mechanisms rely on data structure and embeddings rather than any single human language.
"What Language Does an AI “Think” In? Unpacking How Language Models Process Thought
Large language models don’t “think” like humans—but their internal mechanisms rely on data structure and embeddings rather than any single human language.
By
The Tech Journal
Published about 18 hours ago • 3 min read
When you ask an AI to compose a story in English or translate Italian text into French, it may feel as if the model is “thinking” in that language. But unlike humans, large language models (LLMs) don’t possess consciousness or inner speech—instead, their “thoughts” arise from statistical patterns across massive multilingual datasets. Understanding how they operate can clarify their strengths, limitations, and what truly goes on under the hood.
No Inner Voice: Statistical Patterns, Not Thought
LLMs like GPT, Gemini, Claude, or ChatGPT are neural networks trained on vast volumes of text. They learn to predict the next word or token based on context, building token embeddings and linguistic statistical correlations—not an inner monologue. These embeddings position tokens in a high-dimensional space where semantic and syntactic relationships emerge automatically during training.
Unlike humans, who might think in their native language or even a mixture of languages, LLMs process input purely as mathematical vectors. They don’t shift between languages internally—they handle everything as multilingual patterns learned from data. There is no “hard-coded” English or Italian inside the model—what exists is a web of learned relationships across languages.
Multilingual Harmony: One Universal Space
Many LLMs are trained on multilingual text, enabling sophisticated cross-language capabilities like translation. They do this not by switching between language modules, but by mapping text from different languages into a shared embedding space. For example, the English sentence “Hello, world!” and the Italian “Ciao mondo!” often end up with overlapping internal representations—allowing the model to translate seamlessly.
This shared representation relies on the weak Sapir–Whorf principle, which suggests that language shapes cognition without determining it completely. In LLMs, language shapes their internal geometry but does not confine or define it. This parallels the view of linguists like Boroditsky, who argue that language influences perception but doesn’t strictly limit thought .
Embeddings: The Model’s Memory and “Mind”
Think of word embeddings as the building blocks of an LLM’s “mind.” Each token—whether letter, word, or symbol—gets positioned in a multidimensional space. Training adjusts these positions so embeddings capture meaning and context. Over layers, the model transforms inputs via attention mechanisms, culminating in predictions of next tokens.
When you ask the model a question in Spanish, internally it retrieves embeddings and context that correspond to that input. The entire process stays within the embedding space—there is no literal Spanish voice in its head. Instead, language is patterns in vector space rather than spoken words or inner dialogue.
Key Differences: AI vs. Human Thought
Human thinking often involves inner speech, conceptual representation, and introspection. In contrast, LLMs operate on:
1. Statistical Pattern Recognition: They learn relationships and correlations between tokens but don’t have meaning or intent.
2. Embedding Geometry: Language lives in a vector space representing proximity and relationships between concepts.
3. Multilingual Fusion: Multiple languages get integrated into the same space, enabling translation and cross-lingual reasoning.
4. Black Box Reasoning: While researchers use techniques like attribution graphs to analyze token flow and influence, interpreting how exactly models arrive at decisions remains challenging .
How Researchers Peek Inside the AI “Mind”
Although LLMs are often criticized as “black boxes,” researchers are developing tools to gain insight. Anthropic, for example, uses attribution graphs that visualize how tokens influence each other during generation. These graphs help trace how a model builds a sentence token by token, revealing hidden interactions within the network . While they don’t show inner monologue, they begin to illuminate causal chains of prediction across embeddings and layers.
Practical Consequences of Language Modeling
Since LLMs encode multiple languages in the same space, they can:
-Translate without explicit translation modules.
-Perform cross-lingual tasks like multilingual classification or question answering.
-Exhibit biases based on data representation—ideological, gender, or political biases are captured in embeddings .
While they don’t “think in English” when answering in French, the internal similarity of embeddings ensures coherent, contextually accurate outputs in any supported language.
Final Takeaways
AI language models don’t experience thoughts in human tongues. Instead, they manipulate vectors—mathematical abstractions representing words and concepts across languages. Their “reasoning” is probabilistic and learned across massive datasets, not a silent voice in a particular language.
Still, breaking the model into multilingual embedding spaces and analyzing token flows offers valuable insight into how they process information. As AI advances, a better understanding of embedding structures and interpretability techniques like attribution graphs will be key to assessing reliability, bias, and trustworthiness in language models."
https://vocal.media/01/what-language-does-an-ai-think-in-unpacking-how-language-models-process-thought
#metaglossia_mundus
"China publishes Han-Tibetan version of major dictionary
Source: XinhuaEditor: huaxia2025-07-12 18:32:15
XINING, July 12 (Xinhua) -- China attaches great importance to the use of the Tibetan language, as evidenced by the recent publication of a key bilingual dictionary.
On Friday, the Han-Tibetan version of the Modern Chinese Dictionary, compiled and translated over 13 years, was jointly published by China Tibetology Publishing House and The Commercial Press, according to the ethnic and religious affairs commission of Qinghai Province.
Against the backdrop of rapid social development, Tibetan people have shown a growing desire to learn the national standard Chinese (Han language) while also preserving their own. However, authoritative and user-friendly Han-Tibetan reference books have been scarce, prompting Qinghai -- a province with a large Tibetan population -- to launch the translation project.
The three-volume dictionary has over 70,000 Chinese entries and 200,000 Tibetan entries, totaling more than 10 million characters. While featuring vocabulary, explanations and example sentences with concise, accurate and practical translations, it also integrates word forms and meaning transformations in line with Tibetan grammatical rules.
This dictionary serves as an authoritative language tool in Tibetan-inhabited areas, providing valuable resources for education and cultural exchange and contributing to China's efforts in promoting ethnic exchanges and integration, according to the commission."
https://english.news.cn/20250712/2d5fb1b3d0944f30a24690d868067e9c/c.html
#metaglossia_mundus
"Le Petit Prince prend désormais une nouvelle dimension avec une traduction intégrale en vendéen. Un travail minutieux pour ne pas perdre la poésie du roman original.
Par Romane Rousseau
Publié le 12 juil. 2025 à 14h28
« Crayonne-mu ine ouaille » pour « dessine-moi un mouton »… Le célèbre livre Le Petit Prince, le plus traduit derrière La Bible, se met désormais intégralement au patois vendéen.
Le P’tiot Prince est un projet mené par Jacques Ouvrard, traducteur : « Fin 2024, j’ai écrit Bédame, un ouvrage qui regroupe des termes et expressions en patois vendéen. Nous sommes allés le proposer à la librairie Looc. C’est là qu’on nous a dit que ça pourrait être une bonne idée de traduire Le Petit Prince. »
« Deux univers très différents »
Une tâche parfois rude, le patois vendéen étant un dialecte davantage rural et moins propice à la poésie. « Le Petit Prince a une part de philosophie » explique Claire Pého, éditrice à la Louve des Bois. « C’est tendre, délicat. En le traduisant, il faut garder l’esprit du texte, même avec ce vocabulaire et ces expressions terre-à-terre. On a confronté deux univers très différents. »
Des souvenirs et un héritage
En se replongeant dans ce dialecte, Jacques Ouvrard s’est remémoré de nombreux souvenirs d’enfance, où le patois était commun.
Vidéos : en ce moment sur Actu
« J’ai vécu dans le sud de la Vendée, à la Taillée. À l’école, nous parlions français, mais lorsque l’on jouait dans la cour de récréation, c’était du patois. Avec ma famille, mes grands-parents ou mes cousins, nous parlions également vendéen. C’est très important pour moi de transmettre un peu de cet héritage ».
Une journée dédiée au Petit Prince
Le roman d’Antoine de Saint-Exupéry sera mis à l’honneur le 22 juillet, date de la sortie du livre de Jacques Ouvrard. À l’occasion, Léon Dubois, artiste photographe spécialiste de l’écrivain français, donnera une conférence à 16h à la salle des mariages de la mairie centrale des Sables-d’Olonne.
À 18h, le livre Le P’tiot Prince sera officiellement présenté à la librairie Looc. Jacques Ouvrard sera présent pour expliquer son travail, faire une lecture bilingue (français et vendéen) ainsi que pour dédicacer son œuvre..."
https://actu.fr/pays-de-la-loire/les-sables-d-olonne_85194/une-edition-du-petit-prince-traduite-en-vendeen_62902959.html
#metaglossia_mundus
Words frequently used by ChatGPT, including “delve” and “meticulous,” are getting more common in spoken language, according to an analysis of more than 700,000 hours of videos and podcasts
"ChatGPT Is Changing the Words We Use in Conversation
Words frequently used by ChatGPT, including “delve” and “meticulous,” are getting more common in spoken language, according to an analysis of more than 700,000 hours of videos and podcasts
BY VANESSA BATES RAMIREZ EDITED BY ALLISON PARSHALL
After its release in late 2022, ChatGPT reached 100 million users in just two months, making it the fastest-growing consumer application in history. Since then the artificial intelligence (AI) tool has significantly affected how we learn, write, work and create. But new research shows that it’s also influencing us in ways we may not be aware of—such as changing how we speak.
Hiromu Yakura, a postdoctoral researcher at the Max Planck Institute for Human Development in Berlin, first noticed differences in his own vocabulary about a year after ChatGPT came out. “I realized I was using ‘delve’ more,” he says. “I wanted to see if this was happening not only to me but to other people.” Researchers had previously found that use of large language models (LLMs), such as those that power ChatGPT, was changing vocabulary choices in written communication, and Yakura and his colleagues wanted to know whether spoken communication was being affected, too.
The researchers first used ChatGPT to edit millions of pages of e-mails, essays, and academic and news articles using typical prompts such as to “polish” the text or “improve its clarity.” Next, they extracted words that ChatGPT repeatedly added while editing, such as “delve,” “realm” and “meticulous,” dubbing these “GPT words.” The team then analyzed more than 360,000 YouTube videos and 771,000 podcast episodes from before and after ChatGPT’s release to track the use of GPT words over time. They compared the GPT words with “synthetic controls,” which were formed by mathematically weighting synonyms that weren’t frequently used by the chatbot—such synonyms for “delve,” for example, could include “examine” and “explore.”
The team’s results, posted on the preprint server arXiv.org last week, show a surge in GPT words in the 18 months after ChatGPT’s release. The words didn’t just appear in formal, scripted videos or podcast episodes; they were peppered into spontaneous conversation, too.
“Empirical Evidence of Large Language Model’s Influence on Human Spoken Communication,” by Hiromu Yakura et al. Second version of preprint posted to arXiv.org on June 30, 2025, modified and restyled by Amanda Montañez
“The patterns that are stored in AI technology seem to be transmitting back to the human mind,” says study co-author Levin Brinkmann, also at the Max Planck Institute for Human Development. In other words, a sort of cultural feedback loop is forming between humans and AI: we train AI on written text, it parrots a statistically remixed version of that text back to us, and we pick up on its patterns and unconsciously start to mimic them."
https://www.scientificamerican.com/article/chatgpt-is-changing-the-words-we-use-in-conversation/
#metaglossia_mundus
|
Scoop.it!
