 Your new post is loading...
|
Scooped by
Charles Tiayon
November 7, 2022 10:49 PM
|
Aisha Bewley, a renowned Arabic-English translator, has been named Muslim woman of the year in the 2023 edition of the Muslim 500, a publication that profiles the most influential Muslims. Aisha Bewley [Photo Credit: Facebook] Aisha Bewley, a renowned Arabic-English translator, has been named Muslim woman of the year in the 2023 edition of the Muslim 500, a publication that profiles the most influential Muslims. Ghana, Nigeria contest world’s worst currencies Ms Bewley, who is well known for translating traditional Islamic literatures into English, has worked hard over the years to ensure that the English-speaking Muslim community gets access to Islamic literatures written in Arabic. “Aisha Abdurrahman Bewley (b.1948) is one of the world’s most prolific and accomplished translators of classical Islamic works from Arabic to English,” stated the report. The Muslim 500 publication added that since converting to Islam in 1968, she “has spent the past five decades faithfully learning the Islamic tradition and making its key texts available to the global English-speaking Muslim community, sometimes in collaboration with her husband, Abdalhaqq Bewley, with whom she translated the noble Qur’an.” Married to Hajj Abdalhaqq Bewley, Ms Bewley has written and translated a number of works on Islam to open up the faith to a wider audience because it had previously only been available in Arabic.
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
"Elle relativise l’arrivée de l’IA dans la traduction
Alors que l’IA prétend remplacer les traducteurs...
Clair Pickworth-Guinaudeau, est traductrice indépendante dans le Perche depuis vingt-cinq ans. |
Ouest-France
Publié le 19/08/2025 à 05h01
Trois questions à…
Clair Pickworth-Guinaudeau, traductrice
Comment devient-on traductrice ?
Je viens du Lincoln Shire, au nord de Londres. J’ai suivi une formation traditionnelle en Français et un peu de japonais. J’ai toujours voulu devenir traductrice mais une professeure m’a dit que je devais faire autre chose avant. Ma licence en poche, je suis partie pour six mois en France, d’où je ne suis jamais repartie. J’ai travaillé un temps dans l’enseignement professionnel mais j’avais toujours mon envie de traduction. J’ai donc fait un post graduate à l’Institute of linguistics de Londres et j’ai commencé à travailler de chez moi. Je me rappelle au début les lenteurs d’internet. Je devais attendre trois heures pour télécharger un fichier à traduire ! Pour les jeunes ça doit sembler aussi ringard que de communiquer avec des signaux de fumée… En fait, je suis en lien avec d’autres traducteurs sur des forums ou auprès d’un même client.
Avec l’intellingence artificielle (IA) le métier est-il en danger ?
Certes, les résultats sont impressionnants, il vaut mieux vérifier les textes produits. J’avais un a priori négatif sur l’IA alors dernièrement, j’ai suivi une formation à la Surrey University sur l’IA utilisée pour la traduction. Pour mieux combattre son ennemi, il faut le connaître ! J’utilisais déjà des outils informatisés telles que les mémoires de traduction pour une meilleure cohérence et une mise à jour des traductions successives. Ils gagnent du temps et fidélisent le client. Mais de là à être un métier d’avenir… Seulement en y associant de l’enseignement ou en se spécialisant dans un domaine. La traduction étant en bout de chaîne, les délais sont toujours plus courts. L’IA peut nous permettre de pré-traduire avant de vérifier que la formulation est correcte.
Quels sont vos projets futurs ?
Je vais essayer de varier mes prestations. Je voudrais être traductrice assermentée pour le tribunal car il paraît qu’ils en manquent. Je peux accompagner des personnes qui se préparent à un oral, que ce soit pour un examen ou un exposé professionnel. Mon fils Oliver a une entreprise de production vidéo, nous collaborerons peut-être pour des sous-titrages. Je peux aussi créer une version anglaise de sites internet de gîtes ou d’entreprises pour qu’ils s’ouvrent à d’autres publics. Et aider des Anglais qui veulent acheter ou s’installer en France avec les formalités administratives.
Clair Pickworth-Guinaudeau, partage aussi ses connaissances anglaises en organisant des tea time thématiques à la médiathèque."
https://www.ouest-france.fr/normandie/remalard-en-perche-61110/elle-relativise-larrivee-de-lia-dans-la-traduction-c24ba0c4-7b8d-11f0-86ed-565860524f67
#metaglossia_mundus
"French dictionary gets bad rap over Congolese banana leaf dish
Marthe BOSUANDOLE AFP Aug 14, 2025 Updated Aug 15, 2025
Diners flock to the terrace of Mother Antho Aembe's restaurant in downtown Kinshasa to enjoy "liboke", blissfully unaware of the linguistic brouhaha surrounding the Democratic Republic of Congo's national dish.
Made by grilling fish from the mighty River Congo wrapped in a banana-leaf parcel with spices, tomatoes, peppers, onions, garlic and chillies, liboke enjoys cult status across the central African country.
But liboke's inclusion in one of France's top dictionaries has upset Congolese intellectuals, who say its compilers have failed to capture the full meaning of a word derived from the local Lingala language and closely associated with national identity.
The Petit Larousse dictionary -- an encyclopaedic tome considered a foremost reference on the French language -- announced in May it was including liboke in its 2026 edition.
Its definition: "a dish made from fish or meat, wrapped in banana leaves and cooked over charcoal."
Tucking into a plate on the terrace in the city centre, civil servant Patrick Bewa said it was a "source of pride" that liboke had made it into the leading French dictionary.
"We love it, it's really a typically African and Congolese meal," he said. "With the smoky flavour which takes on the aroma of the leaf, it's an inimitable taste. You have to taste it to believe it."
But some scholars argue that the definition was compiled in Paris by the Academie francaise (French Academy), the chief arbiter on matters pertaining to the French language, without doing justice to liboke's original meanings.
- 'United and undivided' -
Referring only to liboke as food is "very reductive", argued Moise Edimo Lumbidi, a cultural promoter and teacher of Lingala, one of scores of languages spoken in the DRC where French remains the official language.
Under dictator Mobutu Sese Seko, whose rise to power was helped by former colonial master Belgium and whose kleptocratic rule was backed by the United States as a bulwark against Cold War communism, liboke was even part of the national slogan.
"Tolingi Zaire liboke moko, lisanga moko," was a rallying cry, meaning: "We want a united and undivided Zaire", the former name for the DRC during Mobutu's 32 years in power.
"I'm not happy about restricting this precious word, so essential to our culture... liboke moko, it's above all that communion, that national unity," writer and former international cooperation minister Pepin Guillaume Manjolo told AFP.
"Limiting it to its culinary aspects may be all very well for the French, but for us it will not do."
The Petit Larousse should have drawn up the definition by consulting the literary academies of the DRC and its neighbour the Republic of Congo, as the region where the word originated, he said.
AFP contacted the publishers of the Petit Larousse dictionary for comment but did not receive an immediate response.
Edimo, the language teacher, explained that in Lingala, liboke means "a little group".
While liboke's inclusion in the dictionary is a good thing, Edimo said, Larousse's compilers should "deepen their research so as to give us the true etymology of the word".
That would be "a way for them to express their respect for our culture", he added.
At her restaurant in Kinshasa's upscale Gombe district, 41-year-old Mother Aembe was unaware of liboke's newfound literary status, but said she just hoped it would bring in more customers.
mbb/sbk/kjm"
https://www.lebanondemocrat.com/maconcounty/news/national/french-dictionary-gets-bad-rap-over-congolese-banana-leaf-dish/article_44460802-4e64-556b-8ce3-924131888f04.html
#Metaglossia
#metaglossia_mundus
DSPS Specialist/ASL Interpreter
Ventura County Community College DistrictSalary: $82,344.00 - $113,712.00 Annually Job Type: Classified Job Number: 2025-00672 Location: Oxnard College (Oxnard CA), CA Department: Districtwide Closing: 9/25/2025 11:59 PM Pacific Description
WHAT YOU'LL DOUnder the general supervision of an assigned supervisor, a DSPS (Disabled Students Programs and Services)/ASL Interpreter performs a variety of specialized duties involved in the planning, scheduling, and providing of services for students with disabilities; coordinates and implements communication support services for students who are deaf or hard of hearing; serves as an American Sign Language (ASL) interpreter. There is currently one full-time (40 hours/week, 12 months/year) vacancy located at Oxnard College.This recruitment is being conducted to establish a list of eligible candidates that will be used to fill district-wide, current and upcoming, non-bilingual and bilingual, temporary and regular vacancies for the duration of the list, not to exceed one year. WHERE YOU'LL WORKOxnard College was founded in 1975 and is the newest of the three community colleges in the county. Set on 118 acres and located two miles from Pacific Ocean beaches, the college is easily accessible by the Ventura Freeway (Highway 101) or the Pacific Coast Highway.
More information about Oxnard College can be found here: Oxnard College.
WHO WE AREThe Ventura County Community College District (VCCCD) is a public community college district serving residents throughout Ventura County. VCCCD's three colleges - Moorpark College, Oxnard College, and Ventura College - offer programs for transfer to four-year colleges and universities; career technical training, basic skills instruction; as well as community service, economic development, and continuing education for cultural growth, life enrichment, and skills improvement. The Ventura County Community College District recognizes that a diverse community of faculty, staff, and administrators promotes academic excellence and creates an inclusive educational and work environment for its employees, contractors, students, and the community it serves. With the understanding that a diverse community fosters multicultural awareness, promotes mutual understanding and respect, and provides role models for all students, VCCCD is committed to recruiting and employing a diverse and qualified group of administrators, faculty, and staff members who are dedicated to the success of all college students. The Ventura County Community College District does not engage in any employment practice that discriminates against any employee or applicant for employment on the basis of ethnic group identification, race, color, language, accent, immigration status, ancestry, national origin, political beliefs, age, gender, sex, religion, transgender, sexual orientation, marital status, veteran status, and/or physical or mental disability. SALARY PLACEMENT
New Employees: Generally, new employees are placed on the first step of the appropriate range of the salary schedule. Current Employees: An employee who is promoted will be placed on the salary step of the new range of the appropriate salary schedule that provides a minimum increase comparable to a one-step increase in salary. New and current employees may be eligible for advanced step placement as outlined in Section 290 - SALARY PLAN in the Rules of the Personnel Commission for Classified Employees (Download PDF reader).Representative Duties
Serve as a technical resource and provide information and assistance to counselors, and other program faculty and staff regarding the needs and characteristics of individuals who are deaf or hard of hearing, and Deaf culture. EAssess communication styles and ASL language acquirement for students who are deaf or hard of hearing; assign hourly staff compatible with the communication styles of the students; respond to and resolve issues related to hourly staff assignments. ECollaborate with counselors and other program faculty on the development of individual student accommodation plans for students who are deaf or hard of hearing. EInterpret using ASL in classrooms, labs, tutoring sessions, and counseling appointments for students who are deaf or hard of hearing. EAssist counselors and other program faculty with outreach, orientation, and specialized registration assistance for students with disabilities. EProvide information, training, and assistance regarding resources, equipment, supplies, and services available to students who are deaf or hard of hearing; instruct students in the proper operation of specialized software and equipment; arrange for equipment loans. ERecruit, hire, train, provide work direction to, and evaluate hourly sign language interpreters, captioning providers, readers, tutors, scribes, and note takers; maintain work schedules and identify substitutes, as necessary. EProvide input and assistance to program faculty, counselors, and administrators regarding the development and implementation of specialized support services, programs, activities, and projects for students with disabilities. EEnter, retrieve, compile, and organize student data and prepare various reports related to program activities in accordance with State and federal regulations. EMonitor and track hourly staff and captioning services expenses; maintain current budget information. EWork collaboratively and professionally with faculty, staff, students, and stakeholders from diverse academic, socioeconomic, cultural, disability, gender identity, and ethnic communities. EDemonstrate cultural humility, sensitivity, and equity-mindedness in working with individuals from diverse communities; model inclusive behaviors; and achieve equity in assignment-related outcomes. EPerform other duties as assigned. E = Essential Duties Minimum Qualifications
Graduation from high school or evidence of equivalent educational proficiency AND five years of American Sign Language interpreting experience. OR An associate degree from a recognized college or university AND four years of American Sign Language interpreting experience. OR A Bachelor's degree from a recognized college or university AND three years of American Sign Language interpreting experience. LICENSES AND OTHER REQUIREMENTS: Certification from Registry of Interpreters for the Deaf or National Association of the Deaf (Interpreter Level 3 or above) within six months of hire. Supplemental Information
EXAMINATION AND SELECTION PROCESS This is an examination open to the public and current District employees seeking a promotional opportunity. To ensure consideration, please submit your application materials by the posted deadline date on this bulletin. The examination process may consist of any of the following components: A) American Sign Language Skill Evaluation = Qualifying (pass/fail) B) Technical Interview = 100% weighting on final score The candidates with the highest passing scores on the American Sign Language skill evaluation will be invited to the technical interview. The examination process is subject to change as needs dictate. All communication regarding this process will be delivered via email.AMERICAN SIGN LANGUAGE SKILLS EVALUATIONDate Range: Friday, October 3, 2025 to Thursday, October 9, 2025TECHNICAL INTERVIEW
Date Range: Friday, October 17, 2025 to Thursday, October 23, 2025 PLEASE NOTE: The American Sign Language Skill Evaluation will be conducted in person at the District Administrative Center located in Camarillo. The Technical Interviews will be conducted remotely via Zoom.The examination components and dates are subject to change as needs dictate. All communication regarding this process will be delivered via email. SUBMISSION OF APPLICATION Applicants must meet the minimum qualifications as stated, including the possession of licenses, certifications, or other requirements, by the filing deadline in order to move forward in the recruitment process. You must attach copies of any documents that demonstrate your attainment of the minimum qualifications (e.g., unofficial transcripts, foreign transcript evaluation, copies of any required licenses, and/or certifications). Failure to submit any required documents may result in disqualification. All required documentation must be attached to your application; Human Resources staff will not upload your documents for you. The VCCCD does not accept letters of recommendation for classified positions. Please do not attempt to attach letters of recommendation to your application. PLEASE BE AWARE THAT ONCE YOU HAVE SUBMITTED YOUR APPLICATION YOU WILL NO LONGER BE ABLE TO MAKE REVISIONS. If additional versions of your application are submitted, only the most recent will be considered. When completing the application, please make sure you include ALL current and previous employment in the Work Experience section of the application and complete ALL fields, including the name and contact information for your supervisors. Duration of work experience is calculated based off a standard 40-hour full-time work week. Part-time work experience will be prorated based on a 40-hour full-time work week. Experience that is included in the resume but not in the Work Experience section of the application may not be considered for the purpose of determining whether you meet the minimum qualifications. When completing the supplemental questionnaire (if applicable), outline in detail your education, training (such as classes, seminars, workshops), and experience. ELIGIBILITY LIST Upon completion of the examination, the eligibility list will be compiled by combining the final examination score with applicable seniority and veteran's credits, if any. The candidates will be ranked according to their total score on the eligibility list. Certification will be made from the highest three ranks of the eligibility list. This eligibility list will be used to fill current vacancies for up to one year from the date of the technical interview. PROBATIONARY PERIOD All appointments made from eligibility lists for initial appointment or for promotion, with certain exceptions, shall be probationary for a period of six (6) months or one hundred thirty (130) days of paid service, whichever is longer. Classified management, police, and designated executive classifications shall be probationary for a period of one (1) year of paid service from initial appointment or promotion. ACCOMMODATIONS Individuals with disabilities requiring reasonable accommodation in the selection process must inform the Ventura County Community College District Human Resources Department in writing no later than the filing date stated on the announcement. Those applicants needing such accommodation should document this request in an email to HRMail@vcccd.edu including an explanation as to the type and extent of accommodation needed to participate in the selection process. DEGREE INFORMATION If a degree/coursework is used to meet minimum qualifications, an official copy of your transcripts will be required upon hire. If you have a foreign degree and the institution from which your degree was granted is not recognized as accredited by the Council for Higher Education Accreditation (CHEA) or the U.S. Department of Education, foreign transcript evaluation is required if the foreign degree/coursework is used to meet minimum qualifications. The foreign transcript evaluation must be included with your application materials. Visit the Council for Higher Education Accreditation (CHEA) or the U.S. Department of Education to search for institutions that are recognized as accredited. If you need your transcripts evaluated, please review the list of agencies approved for foreign transcript evaluation (Download PDF reader) (Download PDF reader). If applicable, an official copy of your foreign transcript evaluation will also be required upon hire. For more information about the recruitment process at VCCCD, including responses to Frequently Asked Questions, please visit our Classified Careers page. To apply, please visit https://www.schooljobs.com/careers/vcccd/jobs/5036075/dsps-specialist-asl-interpreterjeid-d34de02fe3f87d45b0870de0cbb3b1e2 Send job SavejobClick to add the job to your shortlist
"Abstract: This paper presents the first comprehensive deep learning-based Neural Machine Translation (NMT) framework for the Kashmiri-English language pair. We introduce a high-quality parallel corpus of 270,000 sentence pairs and evaluate three NMT architectures: a basic encoder-decoder model, an attention-enhanced model, and a Transformer-based model. All models are trained from scratch using byte-pair encoded vocabularies and evaluated using BLEU, GLEU, ROUGE, and ChrF + + metrics. The Transformer architecture outperforms RNN-based baselines, achieving a BLEU-4 score of 0.2965 and demonstrating superior handling of long-range dependencies and Kashmiri’s morphological complexity. We further provide a structured linguistic error analysis and validate the significance of performance differences through bootstrap resampling. This work establishes the first NMT benchmark for Kashmiri-English translation and contributes a reusable dataset, baseline models, and evaluation methodology for future research in low-resource neural translation."
Published: 16 August 2025
Deep neural architectures for Kashmiri-English machine translation
Syed Matla Ul Qumar, Muzaffar Azim, …Yonis Gulzar
Scientific Reports volume 15, Article number: 30014 (2025)
https://www.nature.com/articles/s41598-025-14177-8
#Metaglossia
#metaglossia_mundus
"Abstract: Pre-trained models have become widely adopted for their strong zero-shot performance, often minimizing the need for task-specific data. However, specialized domains like medical speech recognition still benefit from tailored datasets. We present ADMEDVOICE, a novel Polish medical speech dataset, collected using a high-quality text corpus and diverse recording conditions to reflect real-world scenarios. The dataset includes domain-specific vocabulary such as drug names and illnesses, with nearly 15 hours of audio from 28 speakers, including noisy environments. Additionally, we release two enhanced versions: one anonymized for privacy-sensitive use and another synthetic version created via text-to-speech, totaling over 83 hours and nearly 50,000 samples. Evaluating the Whisper model, we observe a 24.03 WER on our test set. Fine-tuning with human recordings reduces WER to 15.47, and incorporating anonymized and synthetic data further lowers it to 13.91. We open-source the dataset, fine-tuned model, and code on Kaggle to support continued research in medical speech recognition."
Published: 16 August 2025
A Comprehensive Polish Medical Speech Dataset for Enhancing Automatic Medical Dictation
Andrzej Czyżewski, Sebastian Cygert, …Krzysztof Narkiewicz
Scientific Data volume 12, Article number: 1436 (2025)
https://www.nature.com/articles/s41597-025-05776-1
#metaglossia_mundus
#Metaglossia
"Now We’re Talking: NVIDIA Releases Open Dataset, Models for Multilingual Speech AI
The new Granary dataset, featuring around 1 million hours of audio, was used to train high-accuracy and high-throughput AI models for audio transcription and translation.
Of around 7,000 languages in the world, a tiny fraction are supported by AI language models. NVIDIA is tackling the problem with a new dataset and models that support the development of high-quality speech recognition and translation AI for 25 European languages — including languages with limited available data like Croatian, Estonian and Maltese.
These tools will enable developers to more easily scale AI applications to support global users with fast, accurate speech technology for production-scale use cases such as multilingual chatbots, customer service voice agents and near-real-time translation services. They include:
Granary, a massive, open-source corpus of multilingual speech datasets that contains around a million hours of audio, including nearly 650,000 hours for speech recognition and over 350,000 hours for speech translation.
NVIDIA Canary-1b-v2, a billion-parameter model trained on Granary for high-quality transcription of European languages, plus translation between English and two dozen supported languages. It tops Hugging Face’s leaderboard of open models for multilingual speech recognition accuracy.
NVIDIA Parakeet-tdt-0.6b-v3, a streamlined, 600-million-parameter model designed for real-time or large-volume transcription of Granary’s supported languages. It has the highest throughput of multilingual models on the Hugging Face leaderboard, measured as duration of audio transcribed divided by computation time.
The paper behind Granary will be presented at Interspeech, a language processing conference taking place in the Netherlands, Aug. 17-21. The dataset, as well as the new Canary and Parakeet models, are now available on Hugging Face.
How Granary Addresses Data Scarcity
To develop the Granary dataset, the NVIDIA speech AI team collaborated with researchers from Carnegie Mellon University and Fondazione Bruno Kessler. The team passed unlabeled audio through an innovative processing pipeline powered by NVIDIA NeMo Speech Data Processor toolkit that turned it into structured, high-quality data.
This pipeline allowed the researchers to enhance public speech data into a usable format for AI training, without the need for resource-intensive human annotation. It’s available in open source on GitHub.
With Granary’s clean, ready-to-use data, developers can get a head start building models that tackle transcription and translation tasks in nearly all of the European Union’s 24 official languages, plus Russian and Ukrainian.
For European languages underrepresented in human-annotated datasets, Granary provides a critical resource to develop more inclusive speech technologies that better reflect the linguistic diversity of the continent — all while using less training data.
The team demonstrated in their Interspeech paper that, compared to other popular datasets, it takes around half as much Granary training data to achieve a target accuracy level for automatic speech recognition (ASR) and automatic speech translation (AST).
Tapping NVIDIA NeMo to Turbocharge Transcription
The new Canary and Parakeet models offer examples of the kinds of models developers can build with Granary, customized to their target applications. Canary-1b-v2 is optimized for accuracy on complex tasks, while parakeet-tdt-0.6b-v3 is designed for high-speed, low-latency tasks.
By sharing the methodology behind the Granary dataset and these two models, NVIDIA is enabling the global speech AI developer community to adapt this data processing workflow to other ASR or AST models or additional languages, accelerating speech AI innovation.
Canary-1b-v2, available under a permissive license, expands the Canary family’s supported languages from four to 25. It offers transcription and translation quality comparable to models 3x larger while running inference up to 10x faster.
NVIDIA NeMo, a modular software suite for managing the AI agent lifecycle, accelerated speech AI model development. NeMo Curator, part of the software suite, enabled the team to filter out synthetic examples from the source data so that only high-quality samples were used for model training. The team also harnessed the NeMo Speech Data Processor toolkit for tasks like aligning transcripts with audio files and converting data into the required formats.
Parakeet-tdt-0.6b-v3 prioritizes high throughput and is capable of transcribing 24-minute audio segments in a single inference pass. The model automatically detects the input audio language and transcribes without additional prompting steps.
Both Canary and Parakeet models provide accurate punctuation, capitalization and word-level timestamps in their outputs."
August 15, 2025 by Jonathan Cohen
https://blogs.nvidia.com/blog/speech-ai-dataset-models/
#Metaglossia
"...Some terms commonly used to describe peoples’ interactions with wildlife like “human-wildlife conflict,” “crop-raiding” and “pest” are detrimental to the understanding of animals and their conservation.
“There’s no denying that there will be situations when human and wildlife interests collide, but we can take a step back, consider the power differential between ourselves and other animals, and take a more sympathetic view of these problems,” the author argues.
Language shapes the way we view our world. In the field of wildlife conservation, even very subtle word choices drive peoples’ perceptions around individual species or situations. These word choices can be illustrated by the language associated with wildlife value orientations (WVO).
When asked about views toward wildlife, individuals often fall into one of two camps: domination or mutualism. Those with a domination perspective tend to view wildlife as a resource to be controlled and used according to human needs. People with a mutualism ideology see wildlife as an extended part of our community, and therefore deserving of our respect and protection. Words like “management” or “resources” are associated with a domination perspective, while those with a mutualism orientation are more likely to view wildlife in human terms, or focus on animal welfare.
Wildlife value orientations are influenced by many social and economic factors, and they are not static. Attitudes in both the U.S. and the U.K. have moved from domination to more mutualism orientations, and this change is strongest in younger and more educated groups. As attitudes toward wildlife shift over time, the language we use to describe other animals will naturally change. Conservationists can actively encourage shifting perspectives toward more positive interactions with wildlife by choosing their language carefully. Some terms that should be replaced are “human-wildlife conflict,” “crop-raiding” and “pest,” because this language depicts humans and nonhumans as enemies.
The IUCN defines human-wildlife conflict as “when animals pose a direct and recurring threat to the livelihood or safety of people.” This could mean competition for resources such as food or space, the economic impact of lost crops or livestock, infectious disease transmission, or close proximity to large and physically dangerous wildlife. These issues have been with us for the history of our species, but as the human population continues to grow and require more land, further restricting wild spaces, these issues are becoming more common and more extreme.
It is easy to understand why people would be concerned with these issues. Animals can pose very real threats to peoples’ health, safety and economic stability. However, many scholars have noted that the language of human-wildlife conflict portrays wildlife as if they are actively, purposefully opposed to human interests. One study reviewed scientific papers using that phrase and found that only one of more than 400 studies actually described a conflict situation, in which both actors demonstrate opposing goals or values. This is because wild animals are unaware of human interests and therefore cannot actively work against them. Essentially, the word “conflict” attributes human motivations to nonhuman beings.
“Human-animal conflict” is just one example of wildlife rhetoric evoking crime, war and xenophobia. “Invasive,” “alien” and “foreign” are used to describe animals found outside their expected range. While “invasive” does have an ecological meaning — referring to a species with rapid spread and a detrimental impact on an ecosystem — this is not automatically a result of a species being introduced to a new environment.
Regardless, the animals have not intentionally chosen to invade foreign territory; they are simply trying to survive in a new landscape. One study on urban coyotes describes the importance of language for human-animal relations, finding that depicting coyotes as “invaders” or “foreign” fostered negative opinions of the animals and increased peoples’ fear for their own (and their pets’) safety, despite the fact that the coyotes were residing within their traditional home range...
Tracie McKinney
14 Aug 2025
https://news.mongabay.com/2025/08/its-time-to-update-the-language-of-human-wildlife-interactions-commentary/
#Metaglossia
"Edinburgh University Press has issued directives requiring writers to use capital letters for "Black" while keeping "white" in lowercase when discussing racial matters.
The academic publisher attributes this distinction to what it terms "political connotations".
The publishing house's new guidelines assert that "Black" merits capitalisation as it denotes "a distinct cultural group and a shared sense of identity and community".
Conversely, the instructions explicitly state: "Please do not capitalise 'white' due to associated political connotations".
These requirements appear in the publisher's comprehensive guide on inclusive terminology, which itself carries advisories about "potentially triggering" content.
The publisher's language guidance extends beyond racial terminology to numerous other areas.
Writers must avoid describing migrants as "illegal" and should replace "homeless" with "unhoused" according to the new rules.
Economic terminology faces similar restrictions.
The guide instructs authors to eschew the word "poor" in favour of phrases such as "under-represented", "currently dealing with food insecurity" or "economically exploited".
These stipulations form part of Edinburgh University Press's broader inclusive language framework, which the institution has developed for its annual output of approximately 300 publications...
The guidelines impose restrictions on geographical terminology, forbidding authors from employing broad classifications such as "Eastern" and "Western". Gender-related language faces extensive revision under the new framework.
Writers must eschew terms suggesting binary gender distinctions, including "opposite sex".
The publisher mandates using individuals' chosen pronouns or defaulting to "they" when uncertain. Traditional gendered nouns like "postman" and "chairman" are prohibited.
The guidance concludes with requirements for content warnings...
These mandatory alerts must cover violence, animal cruelty, substance use, transphobia and classism amongst other topics deemed potentially distressing to readers...
These linguistic modifications reflect broader institutional transformations that gained momentum following the 2020 Black Lives Matter demonstrations. Educational establishments have embraced initiatives to confront what they term "embedded whiteness" in professional environments.
The Telegraph disclosed that London Museum personnel received guidance on challenging "whiteness" in their workplace. Teacher training programmes have incorporated modules on disrupting "the centrality of whiteness" within educational settings.
Healthcare institutions have similarly revised their terminology. Various NHS trusts substituted "mother" with expressions like "birthing person" and "people who have ovaries".
Certain services proposed that "chestfeeding" by transgender individuals equates to maternal breastfeeding."
https://www.gbnews.com/news/woke-madness-language-guidance-black-white-ethnicity
#metaglossia_mundus
"Explore how high-quality data and semantic interoperability power the CMS Interoperability Framework and innovation in the health tech ecosystem.
Healthcare technology is set to undergo a monumental transformation, moving from a focus on EHR functionality to a focus on API exchange. At the heart of this evolution lies interoperability, the ability to seamlessly share and use data across different systems. The CMS Interoperability Framework could be emerging as a pivotal force driving this change. But there are other forces at play. By emphasizing shared standards and collaboration among patients, providers, payers, and digital health technologies, these forces are poised to evolve the health tech ecosystem.
Yet, even as some healthcare organizations adopt robust data exchange pathways, the ultimate goal, the last mile as it were, semantic interoperability, remains out of reach for many. While data exchange has improved, the next challenge is making that data complete, meaningful, and actionable...
Understanding the CMS Interoperability Framework
The CMS Interoperability Framework serves as a voluntary roadmap for organizations committed to advancing healthcare data exchange. This framework prioritizes a patient-centered approach and encourages participation from all healthcare stakeholders. By focusing on standardization and enabling technologies, CMS aims to foster seamless collaboration across the health tech ecosystem. Early adopters span the healthcare ecosystem with promises to share data and build patient-facing apps in three topical areas...
...
Operational efficiency
Standardized data helps providers avoid unnecessary repeat tests, missed results, and administrative delays. Semantic interoperability ensures that every piece of information, whether it’s a lab result or a medication list, is complete, accurate, and ready to use. This is essential for initiatives like “Kill the Clipboard,” which aim to eliminate outdated, manual processes in favor of seamless digital workflows...
Terminology management at scale: Centralized terminology management that ensures consistent use of clinical vocabularies like SNOMED CT, LOINC, RxNorm, and ICD-10 across systems. This enables semantic normalization at the point of data ingestion and exchange.
Automated mapping and normalization: Through automated mapping tools, the Health Language platform translates local codes and non-standard terms into standardized terminologies, reducing manual effort and improving data usability for analytics, AI, and patient-facing applications...
Cheryl Mason
Director, Content and Informatics, Health Language
https://www.wolterskluwer.com/en/expert-insights/data-quality-in-the-health-tech-ecosystem
#Metaglossia
L'oeuvre de Cheikh Ahmadou Bamba constitue un patrimoine spirituel et intellectuel d'une richesse inestimable. La traduction et la diffusion des oeuvres de cet homme multidimensionnel et à la piété légendaire contribuent au renforcement des valeurs fondamentales comme le respect, l'amour du travail et la non-violence.
"Sénégal: Traduction et vulgarisation des écrits de Cheikh Ahmadou Bamba - La pensée mouride, passerelle spirituelle entre les cultures
13 AOÛT 2025Le Soleil (Dakar)
Par Aliou DIOUF
L'oeuvre de Cheikh Ahmadou Bamba constitue un patrimoine spirituel et intellectuel d'une richesse inestimable. La traduction et la diffusion des oeuvres de cet homme multidimensionnel et à la piété légendaire contribuent au renforcement des valeurs fondamentales comme le respect, l'amour du travail et la non-violence.
Tous les écrits de Cheikh Ahmadou Bamba sont en arabe. Mais aujourd'hui, ils sont traduits et publiés dans presque toutes les langues vivantes à l'échelle internationale. Et, dans un monde en crise où la montée des tensions et la perte de repères éthiques semblent faire la loi, ces écrits du fondateur du mouridisme pourraient devenir salvateurs. Leur traduction et leur diffusion dans le monde éclairent le chemin des gens vers un futur plus juste et plus pacifique.
Ce, d'autant plus que le message du guide religieux, axé sur la paix, la spiritualité, le travail, le pardon, la patience et la transformation intérieure, a le potentiel de désamorcer bien des tensions et d'ouvrir la voie à une humanité plus fraternelle.
Selon Mouhameth Galaye Ndiaye, théologien et philosophe, la traduction et la diffusion des écrits de Cheikh Ahmadou Bamba à l'échelle internationale revêtent une importance stratégique et spirituelle majeure à l'heure où le monde est en quête de repères éthiques, de paix durable et de dialogue interculturel sincère. Pour lui, traduire et diffuser les écrits de Cheikh Ahmadou Bamba, c'est ouvrir une voie vers un monde « plus juste, plus spirituel et plus solidaire ».
C'est également, poursuit-il, permettre à l'humanité de redécouvrir, dans l'Islam africain, une tradition intellectuelle et mystique d'une puissance méconnue, profondément enracinée dans les valeurs de paix, d'amour du travail, de respect de l'autre et de service désintéressé.
À l'en croire, l'oeuvre du fondateur du mouridisme représente un patrimoine intellectuel et spirituel d'une richesse exceptionnelle, capable d'enrichir aussi bien la pensée islamique que l'humanisme mondial. « L'un des premiers atouts de cette diffusion est qu'elle contribue à consolider les valeurs de paix, de tolérance et de non-violence à l'échelle planétaire », a-t-il expliqué.
Poursuivant son propos, il rappelle que Cheikh Ahmadou Bamba est, à juste titre, considéré comme un apôtre de la non-violence, bien avant des figures mondialement reconnues telles que Gandhi ou Martin Luther King. « Sa résistance à la colonisation française s'est toujours inscrite dans une logique spirituelle de patience, de pardon et de réforme intérieure, et non dans la confrontation armée. Cet héritage, s'il était mieux connu, pourrait nourrir de nouvelles réflexions sur la résolution des conflits et la transformation sociale non violente », affirme-t-il.
Mieux, il estime que la diffusion internationale de l'oeuvre du guide spirituel permettrait l'introduction dans les universités du monde d'une pensée islamique enracinée, rigoureuse et ouverte. Car, d'après lui, Cheikh Ahmadou Bamba ne se contente pas de transmettre des règles religieuses. Il propose une vision intégrale de l'homme et de la société, fondée sur la discipline intérieure, la recherche du savoir, la justice, et surtout le service désintéressé.
« La traduction de l'oeuvre de Cheikh Ahmadou Bamba est d'autant plus justifiée que ce dernier est sans conteste l'un des auteurs les plus prolifiques de l'histoire de l'Islam. Son oeuvre, riche de plusieurs dizaines de milliers de vers, couvre les domaines du droit, de la spiritualité, de l'éthique, de l'histoire, de l'éducation et de la métaphysique », dit-il.
En effet, c'est par la non-violence que Cheikh Ahmadou a choisi de diffuser l'Islam, de résister à la colonisation, mais de protéger son peuple contre les effets dévastateurs de la colonisation et de l'acculturation. C'est cette approche qu'il a proposée pour l'accomplissement du changement social.
Dans sa logique, il ne croyait pas en une réforme de type « top-down », dans laquelle les hommes de pouvoir, laïcs ou religieux, auraient l'initiative. Mais selon lui, pour obtenir un impact durable, il faut que les graines du changement soient semées dans les coeurs et les âmes.
Le type d'éducation qu'il préconisait concernait autant le corps que l'esprit et l'âme. Ainsi, dira Dr Moustapha Diop : « sa philosophie de paix et de pardon peut inspirer le monde, tandis que ses écrits poétiques, appelés "xassaïdes", demeurent des sources d'inspiration spirituelle et culturelle. Imaginez que le monde se conforme à cet enseignement, que deviendraient les conflits et la militarisation ? », s'interroge-t-il avant de rappeler que Cheikh Ahmadou Bamba Mbacké a révélé l'originalité de son approche par l'importance qu'il donne à l'éducation spirituelle qui n'est rien d'autre que la méthode de purification de l'égo.
Cet exercice autour de l'égo, coeur de l'enseignement de Cheikh Ahmadou Bamba, constitue, selon de nombreux spécialistes, un apport majeur au dialogue spirituel mondial. En mettant l'accent sur la transformation intérieure avant toute action extérieure, le fondateur du mouridisme propose un modèle universel applicable à toutes les sociétés.
Son insistance sur l'humilité, la persévérance et la patience dépasse les clivages culturels et religieux. Aujourd'hui, de Dakar à New York, de Touba à Paris, ses « xassaïdes » continuent d'inspirer non seulement les disciples mourides, mais aussi des chercheurs, écrivains et intellectuels curieux de comprendre cette vision pacifique de la réforme sociale.
Plusieurs universités étrangères, notamment aux États-Unis et en Europe, intègrent désormais l'étude des écrits de Cheikh Ahmadou Bamba dans leurs programmes consacrés aux spiritualités africaines et au soufisme.
Lire l'article original sur Le Soleil."
https://fr.allafrica.com/stories/202508130129.html
#metaglossia_mundus
New tech promises to dub film and TV into different languages without losing the quality of the performance.
"Will AI make language dubbing easy for film and TV?
Suzanne Bearne
Technology Reporter
Swedish movie Watch the Skies was dubbed into English using AI
Finding international films that might appeal to the US market is an important part of the work XYZ Films.
Maxime Cottray is the chief operating officer at the Los Angeles-based independent studio.
He says the US market has always been tough for foreign language films.
"It's been limited to coastal New York viewers through art house films," he says.
It's partly a language problem.
"America is not a culture which has grown up with subtitles or dubbing like Europe has," he points out.
But that language hurdle might be easier to clear with a new AI-driven dubbing system.
The audio and video of a recent film, Watch the Skies, a Swedish sci-fi film, was fed into a digital tool called DeepEditor.
It manipulates the video to make it look like actors are genuinely speaking the language the film is made into.
"The first time I saw the results of the tech two years ago I thought it was good, but having seen the latest cut, it's amazing. I'm convinced that if the average person if saw it, they wouldn't notice it - they'd assume they were speaking whatever language that is," says Mr Cottray.
The English version of Watch The Skies was released in 110 AMC Theatres across the US in May.
"To contextualise this result, if the film were not dubbed into English, the film would never have made it into US cinemas in the first place," says Mr Cottray.
"US audiences were able to see a Swedish independent film that otherwise only a very niche audience would have otherwise seen."
He says that AMC plans to run more releases like this.
Writer and director Scott Mann founded the company in 2020, having worked on films including Heist, The Tournament and Final Score.
He felt that traditional dubbing techniques for the international versions of his films didn't quite match the emotional impact of the originals.
"When I worked on Heist in 2014, with a brilliant cast including Robert De Niro, and then I saw that movie translated to a different language, that's when I first realised that no wonder the movies and TV don't travel well, because the old world of dubbing really kind of changes everything about the film," says Mr Mann, now based in Los Angeles.
"It's all out of sync, and it's performed differently. And from a purist filmmaking perspective, a very much lower grade product is being seen by the rest of the world."
Flawless
Scott Mann founded Flawless in 2020
Flawless developed its own technology for identifying and modifying faces, based on a method first presented in a research paper in 2018.
"DeepEditor uses a combination of face detection, facial recognition, landmark detection [such as facial features] and 3D face tracking to understand the actor's appearance, physical actions and emotional performance in every shot," says Mr Mann.
The tech can preserve actors' original performances across languages, without reshoots or re-recordings, reducing costs and time, he says.
According to him, Watch the Skies was the world's first fully visually-dubbed feature film.
As well as giving an actor the appearance of speaking another language, DeepEditor can also transfer a better performance from one take into another, or swap a new line of dialogue, while keep the original performance with its emotional content intact.
Thanks to the explosion of streaming platforms such as Netflix and Apple, the global film dubbing market is set to increase from US$4bn (£3bn) in 2024 to $7.6bn by 2033, according to a report by Business Research Insights.
Mr Mann won't say how much the tech costs but says it varies per project. "I'd say it works out at about a tenth of the cost of shooting it or changing it any other way."
His customers include "pretty much all the really big streamers".
Mr Mann believes the technology will enable films to be seen by a wider audience.
"There is an enormous amount of incredible kind of cinema and TV out there that is just never seen by English speaking folks, because many don't want to watch it with dubbing and subtitles," says Mr Mann.
The tech isn't here to replace actors, says Mann, who says voice actors are used rather than being replaced with synthetic voices.
"What we found is that if you make the tools for the actual creatives and the artists themselves, that's the right way of doing it… they get kind of the power tools to do their art and that can feed into the finished product. That's the opposite of a lot of approaches that other tech companies have taken."
Natan Dvir
Neta Alexander is concerned about a "monolingual" film culture
However, Neta Alexander, assistant professor of film and media at Yale University, says that while the promise of wider distribution is tempting, using AI to reconfigure performances for non-native markets risks eroding the specificity and texture of language, culture, and gesture.
"If all foreign films are adapted to look and sound English, the audience's relationship with the foreign becomes increasingly mediated, synthetic, and sanitised," she says.
"This could discourage cross-cultural literacy and disincentivise support for subtitled or original-language screenings."
Meanwhile, she says, the displacement of subtitles, a key tool for language learners, immigrants, deaf and hard-of-hearing viewers and many others, raises concerns about accessibility.
"Closed captioning is not just a workaround; it's a method of preserving the integrity of both visual and auditory storytelling for diverse audiences," says Prof Alexander.
Replacing this with automated mimicry suggests a disturbing turn toward commodified and monolingual film culture, she says.
"Rather than ask how to make foreign films easier for English-speaking audiences, we might better ask how to build audiences that are willing to meet diverse cinema on its own terms.""
https://www.bbc.com/news/articles/c36xy6r91kwo
#metaglossia_mundus
"Establishing Consensual Terms of AS-OCT Anatomical Landmarks
Anterior-segment optical coherence tomography (AS-OCT) was first described by Izatt et al1 with an OCT wavelength of 830 nm and limited penetration. The advancement from time-domain to spectral-domain to swept-source OCT has led to AS-OCT with longer wavelength (up to 1310 nm), faster scan speed (up to 50 000 A-scans per second), and higher transverse resolution (up to 800 A-scans per B-scan). These improvements provide multiple scans for 360° cross-sectional AS images in high definition. This further allows detailed and repeatable quantitative analysis of various structures, including ocular surface, tear film, cornea, and anterior chamber structures. These advances in AS-OCT have led to numerous publications, advancing our knowledge in pathological structural changes. However, the use of different AS-OCT device models and image processing technologies has resulted in an array of anatomical terminologies, which can be associated with confusion across studies, affecting interstudy interpretability, data comparability, and interpretation of results. To address this problem, the Advised Protocol for OCT Study Terminology and Elements Extension for Anterior Segment (APOSTEL-AS) study group initiated a multinational, multistage consensus exercise to standardize AS-OCT anatomical terminology based on the Accurate Consensus Reporting Document (ACCORD) recommendation."
Poemen P. Chan, MBBS, FCOphthHK1,2,3,4; Julia Y. Chan, MBBS, FCOphthHK, PDip1,2; Carol Y. Cheung, PhD1,3
Published Online: August 14, 2025
doi: 10.1001/jamaophthalmol.2025.2768
https://jamanetwork.com/journals/jamaophthalmology/article-abstract/2837463
A new device could help decode inner speech in paralysis patients – but only on their command – potentially restoring rapid communication.
August 14th, 2025
Scientists develop interface that ‘reads’ thoughts from speech-impaired patients
A new device could help decode inner speech in paralysis patients – but only on their command – potentially restoring rapid communication.
....
The brain’s motor cortex contains regions that control movement – including the muscular movements that produce speech. A BCI uses tiny arrays of microelectrodes (each array is smaller than a baby aspirin), surgically implanted in the brain’s surface layer, to record neural activity patterns directly from the brain. These signals are then fed via a cable hookup to a computer algorithm that translates them into actions such as speech or computer cursor movement.
To decode the neural activity picked up by the arrays into words the patient wants to say, the researchers use machine learning to train the computer to recognize repeatable patterns of neural activity associated with each “phoneme” – the tiniest units of speech – then stitch the phonemes into sentences.
Willett and his colleagues have previously demonstrated that, when people with paralysis try to make speaking or handwriting movements (even though they cannot, because their throat, lip, tongue and cheek muscles or the nerve connections to them are too weak), a BCI can pick up the resulting brain signals and translate them into words with high accuracy.
Recently, the scientists took another important step: They investigated brain signals related to “inner speech,” or language-based but silent, unuttered thought.
Willett is the senior author, and postdoctoral scholar Erin Kunz, PhD, and graduate student Benyamin Meschede-Krasa are the co-lead authors of a new study about this exploration, published Aug. 14 in Cell. (Researchers at Emory University; Georgia Institute of Technology; the University of California, Davis; Brown University; and Harvard Medical School were also involved in the study.)
Willett, the co-director of Stanford’s Neural Prosthetics Translational Laboratory, provided insight on the study’s findings and implications.
What is “inner” speech? And why would a BCI/thought-decoding system that could accurately interpret inner speech be better than one that decodes only attempted speech?
Inner speech (also called “inner monologue” or self-talk) is the imagination of speech in your mind – imagining the sounds of speech, the feeling of speaking, or both. We wanted to know whether a BCI could work based only on neural activity evoked by imagined speech, as opposed to attempts to physically produce speech. For people with paralysis, attempting to speak can be slow and fatiguing, and if the paralysis is partial, it can produce distracting sounds and breath control difficulties.
What did you learn from your efforts to design and employ decoding systems that could discern inner speech?
We studied four people with severe speech and motor impairments who had microelectrode arrays placed in motor areas of their brain. We found that inner speech evoked clear and robust patterns of activity in these brain regions. These patterns appeared to be a similar, but smaller, version of the activity patterns evoked by attempted speech. We found that we could decode these signals well enough to demonstrate a proof of principle, although still not as well as we could with attempted speech. This gives us hope that future systems could restore fluent, rapid, and comfortable speech to people with paralysis via inner speech alone.
Does the system’s potential ability to accurately decode unspoken, silent, inner speech raise issues that hadn’t accompanied previous advances in BCI/decoding software technology?
The existence of inner speech in motor regions of the brain raises the possibility that it could accidentally “leak out”; in other words, a BCI could end up decoding something the user intended only to think, not to say aloud. While this might cause errors in current BCI systems designed to decode attempted speech, BCIs do not yet have the resolution and fidelity needed to accurately decode rapid, unconstrained inner speech, so this would probably just result in garbled output. Nevertheless, we’re proactively addressing the possibility of accidental inner speech decoding, and we’ve come up with several promising solutions.
For people with paralysis, attempting to speak can be slow and fatiguing, and if the paralysis is partial, it can produce distracting sounds and breath control difficulties. ”
It’s worth pointing out that implanted BCIs are not yet a widely available technology and are still in the earliest phases of research and testing. They’re also regulated by federal and other agencies to help us to uphold the highest standards of medical ethics.
What are a couple of the steps that can address this privacy concern?
For current-generation BCIs, which are designed to decode neural activity evoked by attempts to physically produce speech, we demonstrated in our study a new way to train the BCI to more effectively ignore inner speech, preventing it from accidentally being picked up by the BCI. For next-generation BCIs that are intended to decode inner speech directly – which could enable higher speeds and greater comfort – we demonstrated a password-protection system that prevents any inner speech from being decoded unless the user first imagines the password (for example, a rare phrase that wouldn’t otherwise be accidentally imagined, such as “Orange you glad I didn’t say banana”). Both of these methods were extremely effective at preventing unintended inner speech from leaking out.
What lies ahead? How far off is practical realization of this approach? Your next steps?
Improved hardware will enable more neurons to be recorded and will be fully implantable and wireless, increasing BCIs’ accuracy, reliability, and ease of use. Several companies are working on the hardware part, which we expect to become available within the next few years. To improve the accuracy of inner speech decoding, we are also interested in exploring brain regions outside of the motor cortex, which might contain higher-fidelity information about imagined speech – for example, regions traditionally associated with language or with hearing"
https://news.stanford.edu/stories/2025/08/study-inner-speech-decoding-device-patients-paralysis
#metaglossia_mundus
"Language and Cross-Cultural Communication Certificate (Undergraduate)
Admissions
Undergraduate certificate in Language and Cross-Cultural Communication
Acquire the cultural knowledge and sociolinguistic skills necessary to solve language-related problems in everyday and institutional communication with the Undergraduate certificate in Language and Cross-Cultural Communication in Temple University’s College of Liberal Arts. This 15-credit undergraduate certificate is designed to give students the anthropological and linguistic tools needed to understand and study the foundations of language use and communication.
The innovative Language and Cross-Cultural Communication Certificate (Undergraduate) curriculum consists of linguistic anthropology courses that cover language and how it helps us understand culture, contemporary and historical relationships between the two, and the diversity of the world's languages.
Temple University’s overarching goal is to prepare students to work and thrive in an increasingly globalized world. Students with language-related expertise, particularly cross-linguistic and cross-cultural knowledge, have highly desirable skill sets. The Language and Cross-Cultural Communication Certificate (Undergraduate) program is congruent with professional trends in language-related disciplines, and it provides a strong foundation from which to launch a professional career.
Classes & Curriculum
Courses you are likely to take as part of the Language and Cross-Cultural Communication Undergraduate certificate curriculum include
Language and Culture and Theory;
Language, Power, & Identity;
Maya Language and Culture; and
Research Methods in Culture and Communication.
Learn more about Language and Cross-Cultural Communication Certificate (Undergraduate) courses.
Related Graduate Degrees
Africology and African American Studies MA
Africology and African American Studies PhD
Anthropology PhD
Liberal Arts MLA
Spanish MA
Spanish PhD
Tuition & Fees
The tuition for this Graduate Certificate is based on school or college of your major, program degree level (undergraduate or graduate), course load (full- or part-time), in-state or out-of-state residency, and more. You can view the full Cost of Attendance breakdown on the Student Financial Services website. You can also view the 2025–2026 tuition rates."
https://www.temple.edu/academics/degree-programs/language-cross-cultural-communication-certificate-undergraduate-la-lccc-cert
#metaglossia_mundus
"Korea's presidential office launches own sign language interpretation service for press briefings
Viewers of K-drama series may be familiar with the Netflix romance thriller "When the Phone Rings" (2024), featuring a sign language interpreter who works for the presidential office.
While based on a web novel, with a suspense-driven premise seemingly far from reality, the series highlights the importance of communication. Actor Chae Soo-bin stars as Hong Hee-joo, a woman living with selective mutism, who undergoes a highly selective process to be hired as a sign language interpreter for Paik Sa-eon, the presidential spokesperson and her secret husband, played by actor Yoo Yeon-seok.
In a case of reality catching up with fiction, the Korean presidential office is now offering sign language interpretation for press briefings for people with hearing impairments for the first time.
The sign language service was launched Monday at the joint press briefing between President Lee Jae Myung and To Lam, general secretary of Vietnam's Communist Party, following their bilateral summit at the Yongsan presidential office in central Seoul.
The presidential office said that sign language interpretation for briefings is "a measure to guarantee access to information and strengthen social integration and transparency in state administration."
Presidential spokesperson Kang Yu-jung's press briefing Tuesday, which announced Lee's first trip to Washington since taking office for a bilateral summit with U.S. President Donald Trump later this month, was also accompanied by simultaneous sign language interpretation.
"This is the first time that a dedicated sign language interpreter has been hired to provide sign language interpretation support in order to lower the barriers to participation in state affairs for people with hearing and speech impairments," the presidential office said in a statement.
The office said that sign language interpretation will be provided simultaneously for all briefings held in the presidential office briefing room, and that it plans to "gradually expand the scope of interpretation to include major events in the future."
This marks the first time that the presidential office hired dedicated sign language interpreters to provide real-time support for press briefings. Previously, broadcasters usually used their own sign language interpreters when airing presidential press briefings.
The Korean Sign Language Act was enacted in 2016, recognizing Korean sign language as the country's official language of the deaf. The act aimed at improving quality of life and increased participation in society for those with hearing loss.
However, according to a 2023 survey by the National Institute of the Korean Language, 62.9 percent of respondents said there was a need for sign language interpretation in public institutions. Some 86.8 percent of respondents indicated more sign language interpreters were needed in public, financial and medical institutions. These are areas where clear government communication is often most needed.
Public sign language interpretation services have been provided for government announcements and other public sectors since December, but presidential briefings have not been included.
The National Assembly began providing sign language interpretation for press conferences in August 2020.
"We will strive to create a society where every citizen, including deaf people who use sign language as their first language, is not marginalized," the presidential office said."
12 Aug. 2025, 18:24
SARAH KIM
kim.sarah@joongang.co.kr
https://koreajoongangdaily.joins.com/news/2025-08-12/national/socialAffairs/Koreas-presidential-office-launches-own-sign-language-interpretation-service-for-press-briefings/2374142
#metaglossia_mundus
"Microsoft recently released a report ranking translators among the professions most vulnerable to generative AI replacement. In response, ASTA-USA Translation Services, Inc., a leader in human-led translations for over 35 years, underscores for Microsoft and industry leaders the critical limitations of AI in professional translation.
According to Alain J. Roy, Founder and CEO of ASTA-USA Translation Services, Inc., AI is no substitute for professional human translation services, especially for certified, legal, and business-critical conversions. “AI-generated translations may be fast and flashy, but they don’t hold up when you need a legally certified rendering. Period. That’s not a shortcoming of the technology. It’s a non-negotiable legal and regulatory fact,” said Roy.
Based on analysis of over 200,000 anonymized interactions with Bing Copilot, the Microsoft-backed study attempts to measure AI’s relevance across occupations using an “AI applicability score.” Translators, writers, sales reps, and historians are listed as highly susceptible to AI disruption.
According to a July 29 article in Futurism, the report has some obvious faults, such as Microsoft’s natural incentive to put the best spin on generative AI and the lack of a peer-reviewed study. ASTA-USA Translation Services, Inc. cautions that the report overlooks another crucial reality: machine-generated translation isn’t just risky in many industries; it’s inadmissible.
“A financial contract, immigration document, or a clinical trial documentation submitted to regulatory authorities translated by AI alone could jeopardize international deals, trigger regulatory violations, or even land companies in court,” added Roy. “The risk is so significant that in many industries, it’s strictly prohibited.”
While Microsoft’s researchers admit that AI cannot fully perform any one occupation, the broad framing of the study risks oversimplifying the value of human linguists. These translation professionals also interpret context, preserve tone, ensure legal and technical accuracy, and provide certification backed by real credentials. These layers of meaning and accountability are often where AI fails. Not to mention, AI’s hallucinations or omissions can have irreversible consequences.
In an AI-saturated economy, ASTA-USA Translation Services, Inc. proudly highlights its 100% human-performed, multi-layered Quality Assurance process. The company serves leading global corporations, government agencies, and non-profit organizations that require absolute fidelity across language, culture, and compliance frameworks. Their expert translators routinely handle:
Multinational contracts requiring notarized or sworn translation
Legal filings for cross-border litigation
HR documentation and technical manuals with regulatory exposure
Intellectual property documentation and R&D translations
“Unlike AI, our translators carry legal accountability, supported by notarizations, sworn certifications, and recognized professional credentials,” said Roy. “They’re vetted experts, often with industry-specific training in law, medicine, finance, or government.”
As tech companies continue evangelizing AI’s disruptive potential, ASTA-USA Translation Services, Inc., urges decision-makers to separate convenience from compliance. They caution that the illusion of cost-saving can quickly unravel if AI-generated translations lead to lawsuits, regulatory sanctions, or reputation loss.
The fallacy of AI language translation isn’t just theoretical. The translation provider routinely handles emergency corrections after corporations realize an AI engine failed to meet international compliance standards — or worse, mistranslated critical passages entirely.
“Damage control is often far more costly than hiring expert human translators in the first place. AI may have an applicability score,” Roy concludes, “but only human expertise has legal standing.”
...
https://www.digitaljournal.com/pr/news/prunderground/human-translation-services-irreplaceable-asta-usa-1682289043.html
#metaglossia_mundus
"The translation pioneer's new dictionary offers clearer content and a cleaner experience to promote meaningful learning and literacy
NEW YORK, Aug. 12, 2025 /PRNewswire/ -- Reverso, the global leader in online language tools, has expanded its pioneering translation platform with a disruptive new English dictionary for its website and apps: Reverso Define (https://dictionary.reverso.net/). Providing clear, concise definitions for over 500,000 meanings and growing, Reverso Define is built to make it easier for native and non-native users to understand English.
Continue Reading
Reverso Define's clear and clean dictionary entry for the complex noun "culture.”
"Our mission in creating Reverso Define is as clear as its new definitions: to help everyone grow their vocabulary and increase literacy in a fun and simple way," said Théo Hoffenberg, founder and CEO of Reverso. "Reverso Define is reimagining the dictionary as a digital-first companion, delivering quick, rich—and unlike chatbots—consistent information about English words, all just one click away and easily saved for deeper learning."
Combining the rigor of linguistics, power of data, best practices in education, and assurance of meticulous review, Reverso Define provides a host of AI-powered features:
Shorter definitions illustrated with real-world examples, context labels, and synonyms for every meaning
Comprehensive coverage: over 100,000 expressions and idioms, 5,000 abbreviations, and extensive slang, technical language, and more
Instant results for 98% of lookups, including inflected forms and phrases as headwords
Clickable links on every word in definitions for effortless exploration
Pronunciations for every definition, meaning, and example in American and British English
Images for 50,000+ meanings, especially high-frequency concrete nouns
Regular revision, with approximately 10,000 meanings upgraded each update
Immediate, cross-referenceable translation in Reverso's 25+ languages
Developed by Reverso's team of language specialists, data scientists, and engineers and backed by the company's nearly 30 years of innovation in language technology, Reverso Define offers a clean experience seamlessly integrated into Reverso's all-in-one language ecosystem.
Perfect for English as both a first and foreign language, Reverso Define enables teachers to easily create thematic and leveled word lists, which students can study using quizzes, flashcards, and multimodal listening and pronunciation practice tools. With one tap, everyday users get short and sweet definitions in a snap—whether reading articles, writing, emailing, texting, or watching videos online.
"Most of today's electronic dictionaries have evolved from traditional dictionaries. They were not created specifically for on-screen reading," said Dr. Ana Frankenberg-Garcia, member of the editorial board of Oxford Academic's International Journal of Lexicography and longtime contributor to renowned dictionaries. "Reverso is a new English dictionary designed to help you understand unfamiliar words and expressions with minimal disruption while concentrating on what you are reading."
Reverso Define is free and now available on the web (https://dictionary.reverso.net/) and app (https://context.reverso.net/translation/mobile-app/) as well as through Reverso's desktop downloads and browser extensions.
In June, Reverso released a new version of its award-winning app, Reverso Translate and Learn, the first of its kind able to be set as the default translator in devices running iOS 18.4 and later.
About Reverso
Reverso is a global leader in online translation and language tools, helping millions of people and professionals read, write, learn, and communicate across the world's languages. For over 100 language combinations, Reverso provides AI-powered contextual voice, image, text, and document translation and an all-in-one learning ecosystem with a top grammar checker and user-first English dictionary. Each month, it serves over 50 million active users on the web, 5 million app users, and 5 million users in corporate environments.
Press Contact
John Kelly
Boldsquare
john.kelly@boldsquare.com
SOURCE Reverso"
Aug 12, 2025, 09:00 ET
https://www.prnewswire.com/news-releases/reverso-launches-reverso-define-an-english-dictionary-built-for-the-modern-user-and-learner-302527245.html
#metaglossia_mundus
"Des chercheurs autochtones de différents coins du monde se servent de l’intelligence artificielle (IA) pour répondre aux besoins spécifiques de leurs nations et démontrent qu’elle peut être utilisée même dans des contextes culturels très éloignés de celui qui l’a vue naître.
« Après la génération qui a aujourd’hui entre 80 et 90 ans, il y aura très peu de personnes ayant grandi dans un environnement entièrement autochtone sur le plan linguistique. C’est donc crucial de préserver ce savoir », explique Jason Edward Lewis, professeur d’arts numériques à l’Université Concordia d’origine hawaïenne et samoane, qui envisage l’IA comme un moyen d’éviter la perte de ces héritages.
M. Lewis, qui s’attelle depuis 20 ans à intégrer les membres des premiers peuples dans les espaces virtuels, est également codirecteur d’Abundant Intelligences, un projet de recherche qui vise à développer des systèmes d’IA en collaboration avec six communautés au Canada, aux États-Unis et en Nouvelle-Zélande. Il était présent au premier rassemblement mondial sur les IA autochtones de l’Institut québécois d’intelligence artificielle Mila à la mi-juillet.
Cercles de parole, confection de bourses sacrées et chants traditionnels y accompagnaient les présentations d’experts internationaux animées autour de la pluralité d’usages et de méthodes proposés. Dans la lignée de la mission d’Abundant Intelligences, l’objectif de ce ralliement visait à « soutenir les connaissances ancestrales » et à « reconnaître la multiplicité de façons d’être intelligent dans le monde ».
C’est ce qu’ont tenté de faire les chercheurs à l’origine du projet Buffalo In Motion, par exemple, qui aspire à réintroduire le bison sur les terres ancestrales anichinabées en utilisant « les données de migration issues du suivi GPS et les données écologiques pour observer les bénéfices écologiques de leur retour ». Idem pour ceux qui sont à l’origine de la plateforme SAIGE, qui centralise la demande de bourses universitaires pour les candidats autochtones afin de les inciter à rester plus longtemps sur les bancs d’école en améliorant leur accès au financement.
Ces deux projets sont issus du programme « Éclaireurs autochtones en IA », une initiative pilotée par l’institut Mila et l’organisation Indspire, qui réunit depuis l’an dernier des chercheurs issus des Premières Nations, des Inuits et des Métis d’un bout à l’autre du Canada. Cette résidence d’été cible des candidatures autochtones afin d’inciter les nouveaux talents « à apprendre, à développer et à diriger l’évolution de l’IA ».
Un défi complexe
Michael Running Wolf, d’ascendance lakota, cheyenne et pied-noir, est cofondateur et architecte du projet appliqué de Mila First Languages AI Reality (FLAIR), qui revitalise les langues autochtones en recourant à l’IA et aux technologies immersives pour dynamiser leur enseignement. Un défi de taille, puisque plusieurs d’entre elles sont « polysynthétiques », à savoir que chacun de leurs mots combine de nombreux éléments de sens qui dépendent du contexte et de l’environnement immédiat dans lequel il est prononcé. M. Running Wolf cite l’exemple du mot « voiture », qui pourrait, selon la situation, devenir « la-voiture-rouge-au-nord-de-chez-moi-qui-appartient-à-mon-frère ».
Cette étendue de mots pratiquement infinie devient alors très complexe à intégrer aux principaux modèles d’IA qui traitent le langage, entraînés à reconnaître les langues indo-européennes.
À l’instar de Jason Edward Lewis, Michael Running Wolf ne croit pas que l’IA soit la formule magique qui permettra de « sauver » les langues autochtones d’une disparition certaine, bien qu’elle représente un véhicule utile à cet égard. Il appelle à une « réappropriation » des outils de l’IA pour collecter des données « et ensuite les utiliser pour revitaliser les langues », plutôt que de seulement les « préserver ».
Pallier la sous-représentation
Les communautés autochtones sont quasiment absentes du virage numérique et du paysage actuel de l’IA, bien qu’il existe quelques figures de proue. « Elles représentent environ 5 % de la population, mais environ 1 % de la main-d’œuvre numérique, et on estime que ce chiffre est inférieur dans le domaine de l’IA », souligne Lynnsey Chartrand, responsable des initiatives autochtones, des politiques publiques et de l’inclusion à l’institut de recherche Mila.
Le constat d’Elissa Strome, directrice générale de la stratégie pancanadienne en matière d’IA à l’Institut canadien de recherches avancées (CIFAR), est qu’il y a encore du pain sur la planche avant d’atteindre une certaine équité. « Il faut que ce débat soit beaucoup plus large, dans les universités et les laboratoires de recherche du pays. Je n’ai vu aucun exemple de grandes entreprises s’intéresser aux perspectives autochtones en IA. » Par contre, celle-ci observe depuis environ deux ans un nombre croissant de startups dirigées par des Autochtones dans ce domaine.
Une « tendance prometteuse » encouragée par M. Running Wolf, qui a grandi dans un village rural du Montana, où l’eau et l’électricité étaient intermittentes. « Créer une startup dans le Nord-du-Québec est difficile, car il n’y a pas d’Internet. Mais si vous commencez à dire : “Je vais créer une startup d’IA dans le Nord-du-Québec, mais il me faut d’abord Internet”, c’est facile à vendre. »"
Mariane Laporte
https://www.ledevoir.com/societe/science/908782/polyculture-ia-autochtones
#metaglossia_mundus
"The San Francisco–based Center for the Art of Translation (CAT) will hold its 2025 Day of Translation Festival at Brooklyn’s Center for Fiction on September 18.
The event includes three free, livestreamed panels from noon to 5 p.m.: “Resistance Translation,” a reflection on how translation can help subvert dominant narratives; “Meeting the Present Moment,” on translators navigating cultural taboos and censorships; and “Interspecies Translation,” highlighting the languages of the animal and natural worlds.
Among the panelists are literary translators Susan Bernofsky, Anton Hur, and Chenxin Jiang. The day will culminate in a ticketed keynote conversation featuring novelists Jhumpa Lahiri and Katie Kitamura.
Now in its sixth year, the Day of Translation Festival “connects readers of literary translation; literary translators at every stage of their careers; and anyone interested in the movement of ideas among languages, cultures, people, and places,” per the announcement from CAT, which is also home to Two Lines Press.
Registration is open now on the CAT website."
https://www.publishersweekly.com/pw/newsbrief/index.html?record=5489
#metaglossia_mundus
Straker says its small language models offer faster, cheaper, and more accurate translations by focusing on industry-specific context rather than general language coverage.
"Small language models gain ground in AI translation
Straker says its small language models offer faster, cheaper, and more accurate translations by focusing on industry-specific context rather than general language coverage.
Small language models are emerging as a serious challenger to large, general-purpose AI in translation, offering faster turnaround, lower costs, and greater accuracy for specific industries and language pairs.
Straker, an ASX-listed language technology firm, claims its Tiri model family can outperform larger systems by focusing on domain-specific understanding and terminology rather than broad coverage.
Tiri delivers higher contextual accuracy by training on carefully curated translation memories and sector-specific data, cutting the need for expensive human post-editing. The models also consume less computing power, benefiting finance, healthcare, and law industries.
Straker integrates human feedback directly into its workflows to ensure ongoing improvements and maintain client trust.
The company is expanding its technology into enterprise automation by integrating with the AI workflow platform n8n.
It adds Straker’s Verify tool to a network of over 230,000 users, allowing automated translation checks, real-time quality scores, and seamless escalation to human linguists. Further integrations with platforms like Microsoft Teams are planned.
Straker recently reported record profitability and secured a price target upgrade from broker Ord Minnett. The firm believes the future of AI translation lies not in scale but in specialised models that deliver translations that are both fluent and accurate in context."
12 Aug 2025
https://dig.watch/updates/small-language-models-gain-ground-in-ai-translation
#metaglossia_mundus
"OpenAI's GPT-5: India may become our largest market, says CEO Sam Altman
India, currently OpenAI's second-largest market, could soon become its largest globally, CEO Sam Altman said on Wednesday as the company unveiled its next-generation model, GPT-5, which will be available free to all users.
Hailing India as an "incredibly fast-growing" market, Altman noted the remarkable pace at which Indian citizens and businesses are adopting Artificial Intelligence (AI).
"India is our second-largest market in the world after the US, and it may well become our largest market. It's incredibly fast-growing, but what users are doing with AI, what citizens of India are doing with AI, is really quite remarkable.
"We're especially focused on bringing products to India, working with local partners to make AI work great for India and make it more affordable for people across the country.
"We've been paying a lot of attention here given the rate of growth and I am excited to come for a visit in September," Altman said during a media briefing.
OpenAI on Thursday announced the launch of GPT-5, describing it as its "best model yet for coding and agentic tasks".
"We're releasing GPT-5 in three sizes in the API -- gpt-5, gpt-5-mini, and gpt-5-nano -- to give developers more flexibility to trade off performance, cost, and latency. While GPT-5 in ChatGPT is a system of reasoning, non-reasoning, and router models, GPT-5 in the API platform is the reasoning model that powers maximum performance in ChatGPT.
"Notably, GPT-5 with minimal reasoning is a different model than the non-reasoning model in ChatGPT, and is better tuned for developers. The non-reasoning model used in ChatGPT is available as gpt-5-chat-latest," OpenAI said in a blogpost.
The new model is a major upgrade over its predecessor, GPT-4, and represents a "pretty significant step" toward Artificial General Intelligence (AGI), Altman said.
"GPT-5 is really the first time that I think one of our mainline models has felt like you can ask a legitimate expert, like a PhD-level expert, anything... We wanted to simplify it and make it accessible. We wanted to make it available in our free tier for the first time," he said.
Nick Turley, head of ChatGPT, added that the new model significantly improves understanding across more than 12 Indian languages.
Whatsapp Banner"GPT-5 significantly improves multilingual understanding across over 12 Indian languages, including regional languages. So that's really exciting because as Sam mentioned, India is a priority market for us," Turley said."
OpenAI's GPT-5: India may become our largest market, says CEO Sam Altman - The Economic Times"
By PTI
Aug 08, 2025, 10:18:00 AM IST https://share.google/QiXfkOch9S3dqh9c4
#metaglossia_mundus
"Use of Large Language Models for Extracting and Analyzing Data from Heterogeneous Catalysis Literature
Abstract: Extracting experimentally measured heterogeneous catalysis data from the text of research articles into structured databases would facilitate the rapid screening of catalysts with target properties and the development of models capable of directly predicting experimental outcomes. This text mining task has been transformed by the release of large language models (LLMs) capable of following general natural language instructions, which have made it possible to mine text without the need to train task-specific models or define comprehensive expression-matching rules. Here, we develop and share a text mining tool called CatMiner that extracts arbitrary user-specified structure–environment–property data using LLMs. It is agnostic to LLM choice, with both OpenAI GPT models and open-source Llama and DeepSeek models supported without modification. We benchmark the ability of CatMiner to rapidly extract useful data from abundant published literature by focusing on a case study of the oxidative coupling of methane. We explore how model choice and prompting strategies affect extraction quality. Key capabilities, including the use of domain knowledge, iterative prompting, and document-wide context handling are shown to be critical for effective performance. We identify situations where CatMiner struggles and suggest reporting standards for the community to make catalysis data easier to extract going forward. CatMiner enables the creation of machine-readable catalysis datasets, streamlining access to experimental insights buried in the literature.
Benjamin W. Walls, Suljo Linic*
Cite this: ACS Catal. 2025, 15, XXX, 14751–14763
https://doi.org/10.1021/acscatal.5c03844
Published August 10, 2025
© 2025 The Authors. Published by American Chemical Society
https://pubs.acs.org/doi/10.1021/acscatal.5c03844
#metaglossia_mundus
"L'IA dans l’éducation africaine : progrès ou perte de mémoire culturelle ?
Les outils d’IA dans l’éducation africaine risquent d’effacer les valeurs autochtones ; une maîtrise locale et une IA ancrée dans la culture sont essentielles pour préserver les savoirs et garantir l’équité.
11 août 2025
Par James Maisiri et Solomon Musonza
Ces dernières années, un discours dominant s’est imposé, prônant l’intelligence artificielle (IA) comme moyen d’améliorer les systèmes éducatifs. À ce titre, le secteur éducatif africain utilise de plus en plus des outils pédagogiques basés sur l’IA, avec la promesse que ces technologies peuvent aider les enseignants dans leurs tâches quotidiennes, offrir un apprentissage personnalisé aux élèves et améliorer les pratiques pédagogiques. Cependant, une dimension importante de ce débat est souvent négligée : la technologie n’est pas neutre. Elle promeut, restreint et façonne les cultures de ceux qui la reçoivent. Peut-être devrions-nous, dans les systèmes éducatifs africains, réfléchir de manière critique aux valeurs et aux visions du monde qui se reflètent dans la conception des outils éducatifs basés sur l’IA.
De nombreux chercheurs ont conclu que la technologie n’est pas neutre, mais qu’elle représente souvent intrinsèquement les idéologies, les pratiques culturelles et les valeurs politiques du contexte dans lequel elle a été conçue (Van de Poel & Kroes, 2014). Par exemple, des féministes ont soutenu que les cultures patriarcales ont façonné la conception de technologies selon une perspective masculine (Hipolito et al., 2023). Parmi les exemples figurent des assistants virtuels comme Siri et Alexa, conçus comme des assistantes dociles, amicales, voire légèrement séductrices, à la voix féminine, exécutant toutes les instructions, perpétuant ainsi des stéréotypes patriarcaux (Strengers & Kennedy, 2021). Au Japon, certaines technologiques reflètent les croyances spirituelles shintoïstes selon lesquelles les objets non humains ont une âme, comme on le voit avec les animaux de compagnie robotiques alimentés par l'IA, tel le chien AIBO de Sony, créé pour refléter les valeurs sociétales de la compagnie (Jensen, 2013). Les digues maritimes et les dos d'âne sont également des technologies conçues pour incarner la valeur sociétale de la protection de la vie humaine (Van de Poel & Kroes, 2014).
Peut-être devrions-nous, dans les systèmes éducatifs africains, réfléchir de manière critique aux valeurs et aux visions du monde qui se reflètent dans la conception des outils éducatifs basés sur l’IA.
Si nous acceptons que la technologie puisse façonner et incarner intrinsèquement des valeurs culturelles, cela signifie qu’un transfert de technologie est aussi un transfert des valeurs culturelles de ses concepteurs (Majumder & Tripathi, 2020). Dans le domaine de l’éducation, le projet One Laptop Per Child (OLPC), mis en œuvre dans plusieurs pays, a bouleversé l’apprentissage communautaire traditionnel au Nigeria et ailleurs en promouvant un enseignement individualisé et autodirigé, en contradiction avec les valeurs locales (Ezumah, 2020). Étant donné que les outils éducatifs basés sur l’IA proviennent souvent des contextes culturels des pays du Nord et sont de plus en plus utilisés dans l’éducation en Afrique, il faut se demander comment ces valeurs et ces préjugés intégrés seront perçus.
Nyaaba, Wright et Choi (2024) soutiennent que les outils éducatifs d’IA sont conçus pour produire des programmes et des contenus destinés aux pays occidentaux technologiquement avancés, ignorant fréquemment les coutumes et savoirs autochtones. Par exemple, dans de nombreuses sociétés africaines, les systèmes éducatifs incluent des pratiques agricoles fondées sur des connaissances écologiques contextuelles. Or, ce savoir est souvent absent des programmes générés par l’IA. Ainsi, les élèves qui utilisent l’IA reçoivent un transfert culturel de méthodes externes tout en étant privés de contenus d'apprentissage autochtones. Cette tendance à l’effacement épistémique des méthodes agricoles africaines n’est pas nouvelle et a été historiquement introduite par l’impérialisme et la colonisation à travers les processus d’industrialisation, mais l’IA risque de contribuer encore davantage à l’effacement des savoirs autochtones. Par exemple, la rationalité occidentale perçoit les plantes comme des objets, alors que les connaissances autochtones d’Afrique de l’Est considèrent les plantes comme des êtres dotés d’esprit et d'une capacité d'action. Or, ces valeurs Est-Africaines sont souvent absentes des programmes agricoles générés par l’IA (Foster et al., 2023). Par ailleurs, lorsque des chercheurs ont demandé à ChatGPT et Gemini combien il y a de saisons, les deux ont répondu « quatre ». En Afrique de l’Ouest, cependant, il existe deux saisons principales, ce qui montre que ces outils ont tendance à trop généraliser les variations climatiques et peinent à fournir des données éducatives précises et contextualisées (Nyaaba, Wright et Choi, 2024).
De plus, de nombreux systèmes éducatifs africains autochtones sont axés sur la communauté. Par exemple, l’éducation autochtone des peuples Massaï et Kipsigis au Kenya met l’accent sur les compétences interpersonnelles et sociales comme éléments essentiels de la culture (Kerubo, 2016). Cependant, nombre de technologies éducatives sont conçues autour de valeurs liées à l’apprentissage individualisé. Des outils comme ChatGPT favorisent également un apprentissage individualisé et centré sur l’écran, ce qui menace une culture d’apprentissage communautaire, collaboratif et interpersonnel. Certaines études ont même observé une corrélation entre l’utilisation de l’IA par les élèves à l’école et une diminution de la capacité d'adaptation sociale des élèves (Lai et al., 2023 ; Kouam & Muchowe, 2025).
Au-delà des programmes scolaires, l’appropriation de la langue représente également une menace pour les valeurs culturelles. Les langues ne peuvent être réduites à de simples moyens de communication : elles incarnent un cadre épistémologique spécifique. Les outils éducatifs d’IA sont multilingues mais souvent monoculturels, conservant les préjugés culturels de la langue dans laquelle ils ont été entraînés, excluant en grande partie les épistémologies et langues autochtones. Par exemple, dans certaines communautés africaines, la spiritualité occupe une place importante dans le programme éducatif. Pourtant, lorsqu’on a demandé à Gemini de proposer une formulation différente de « Dieu est bon » en gurune, une langue parlée au Ghana, la réponse a révélé sa faible maîtrise de cette langue et son manque de connaissances à son sujet (Nyaaba, Wright et Choi, 2024). En incitant les étudiants africains à se conformer au discours occidental et aux visions du monde qu’il véhicule, plutôt qu’à valoriser leurs langues et systèmes de savoirs autochtones dans les outils d’IA, on constate un transfert culturel au sein de leur éducation.
Étant donné que les outils éducatifs basés sur l’IA proviennent souvent des contextes culturels des pays du Nord et sont de plus en plus utilisés dans l’éducation en Afrique, il faut se demander comment ces valeurs et ces préjugés intégrés seront perçus.
Les écosystèmes numériques créent des « territoires numériques qui, à l’instar des espaces physiques, ont la propension à devenir des lieux… de colonialité numérique-territoriale » (Mohamed et al., 2020 : 65). Les systèmes éducatifs africains ont largement importé des outils d’IA sans se soucier des conséquences épistémologiques. La dépendance technologique de l’Afrique vis-à-vis des pays du Nord la rend vulnérable à un transfert unilatéral de valeurs et de cadres culturels, entraînant un déplacement culturel. Les établissements d'enseignement et les gouvernements africains doivent créer leurs propres outils éducatifs d’IA, qui reflètent les systèmes de savoirs autochtones, tout en éduquant activement leur population dans les domaines des sciences, de la technologie, de l’ingénierie et des mathématiques (STEM) afin de rompre cette dépendance technologique vis-à-vis des pays du Nord et de forger leur propre destin.
En incitant les étudiants africains à se conformer au discours occidental et aux visions du monde qu’il véhicule, plutôt qu’à valoriser leurs langues et systèmes de savoirs autochtones dans les outils d’IA, on constate un transfert culturel au sein de leur éducation.
Au-delà des réformes descendantes, il existe en Afrique des mouvements populaires qui luttent contre les préjugés liés à l'IA et l'injustice épistémique dans leur contexte éducatif. Par exemple, Masakhane est une initiative panafricaine bénévole qui met en place un réseau de traitement du langage naturel (NLP) formé par des étudiants, des enseignants, des chercheurs et des codeurs africains. Cette initiative vise à traduire les informations disponibles en ligne dans les langues maternelles d’apprenants issus de divers horizons (Ravindran, 2023). Ils ont lancé l'initiative « Décoloniser la science », qui visait initialement à traduire des articles de recherche pré-publiés dans le domaine des STIM en plusieurs langues africaines. Dans un premier temps, ils ont commencé par traduire les résumés de ces articles dans différentes langues africaines. Les systèmes d’IA de Masakhane sont souvent créés grâce à une recherche participative favorisant la collaboration ouverte via des contributions communautaires, par exemple sous la forme d'enregistrements de discussions communautaires ou d'articles de journaux locaux (Ravindran, 2023). Ce processus de recherche participative démocratise l’IA et encourage l’appropriation locale de la production de savoirs autochtones.
La présence d’initiatives locales, aux côtés de stratégies nationales, peut aider les systèmes éducatifs africains à créer des solutions d’IA équitables et culturellement pertinentes. Au-delà des solutions de formation à l’IA, une utilisation pertinente de l’IA dans les systèmes éducatifs africains doit impérativement s’accompagner d’une appropriation des outils éducatifs d’IA. Les systèmes éducatifs africains doivent se doter des moyens de décider quand et comment utiliser ces outils dans le contexte africain.
References
Ezumah, B. A. (2020). Critical Perspectives of Educational Technology in Africa, Design, Implementation, and Evaluation. Springer. https://link.springer.com/book/10.1007/978-3-030-53728-9 (Accessed: 25 October 2024).
Foster, L., Szilagyi, K., Wairegi, A., Oguamanam, C., & de Beer, J. (2023). Smart farming and artificial intelligence in East Africa: Addressing indigeneity, plants, and gender. Smart Agricultural Technology, 3, 100132. https://doi.org/10.1016/j.atech.2022.100132 (Accessed: 25 January 2025).
Hipolito, I., Winkle, K., & Lie, M. (2023). Enactive Artificial Intelligence: Subverting Gender Norms in Robot-Human Interaction. https://www.frontiersin.org/journals/neurorobotics/articles/10.3389/fnbot.2023.1149303/full (Accessed: 24 February 2025).
Jensen, C. (2013). Techno-animism in Japan: Shinto Cosmograms, Actor-network Theory, and the Enabling Powers of Non-human Agencies. Theory, Culture & Society, 30(2), 84–115.
Kouam, A. W. F., & Muchowe, R. M. (2024). Exploring graduate students’ perception and adoption of AI chatbots in Zimbabwe: Balancing pedagogical innovation and development of higher-order cognitive skills. Journal of Applied Learning and Teaching, 7(1), 1-11. https://dx.doi.org/10.2139/ssrn.4748274
Kerubo, I. P. (2016). African indigenous education as practiced by the Maasai of Kenya [Master's thesis, University of Nairobi]. University of Nairobi eRepository. https://shorturl.at/uSFIL (Accessed: 14 December 2024).
Lai, T., Xie, C., Ruan, M., Wang, Z., Lu, H., & Fu, S. (2023). Influence of artificial intelligence in education on adolescents' social adaptability: The mediatory role of social support. PLoS One, 18(3), e0283170. https://doi.org/10.1371/journal.pone.0283170
Majumder, M., & Tripathi, A. K. (2020). Transformative power of technologies: Cultural transfer and globalization. AI & Society, 38, 2295–2303. https://doi.org/10.1007/s00146-021-01144-w (Accessed: 15 April 2025).
Mohamed, S., Png, M.-T., & Isaac, W. (2020). Decolonial AI: Decolonial Theory as Sociotechnical Foresight in Artificial Intelligence. Philosophy & Technology, 33, 659–684.
Nyaaba, M. et al. (2024) Generative AI and Digital Neocolonialism in Global Education: Towards an Equitable framework, [Preprint] arXiv. journal-article, https://arxiv.org/abs/2406.02966 (Accessed: 25 January 2025).
Ravindran, S. (2023). ‘AI often mangles African languages. Local scientists and volunteers are taking it back to school.’ Science. https://doi.org/10.1126/science.adj8509. (Accessed: 15 May 2025).
Strengers, Y., & Kennedy, J. (2020). The smart wife: Why Siri, Alexa, and other smart home devices need a feminist reboot. Mit Press. https://mitpress.mit.edu/9780262542791/the-smart-wife/ (Accessed: 15 January 2025).
van de Poel, I., & Kroes, P. (2014). Can Technology Embody Values? In P. Kroes & P.-P. Verbeek (Eds.), The Moral Status of Technical Artefacts (pp. 103–124). Springer Netherlands. https://doi.org/10.1007/978-94-007-7914-3_7 (Accessed: 17 November 2024).
Les idées exprimées ici sont celles des auteurs ; elles ne reflètent pas nécessairement la position officielle de l'UNESCO et n'engagent pas l'Organisation.
Biographies
James Maisiri est chercheur et doctorant à l’Université de Johannesburg, où il étudie la manière dont les technologies numériques transforment les marchés du travail dans un contexte africain, avec des implications pour l’éducation et le développement des compétences.
Solomon Musonza est stagiaire de recherche au Centre for Social Development in Africa (CSDA) de l’Université de Johannesburg, où il se concentre sur l’utilisation de ChatGPT par les étudiants dans un cadre éducatif."
https://www.unesco.org/fr/articles/lia-dans-leducation-africaine-progres-ou-perte-de-memoire-culturelle
#metaglossia_mundus
|
"Aisha Bewley, a renowned Arabic-English translator, has been named Muslim woman of the year in the 2023 edition of the Muslim 500, a publication that profiles the most influential Muslims.
Ms Bewley, who is well known for translating traditional Islamic literatures into English, has worked hard over the years to ensure that the English-speaking Muslim community gets access to Islamic literatures written in Arabic.
“Aisha Abdurrahman Bewley (b.1948) is one of the world’s most prolific and accomplished translators of classical Islamic works from Arabic to English,” stated the report.
The Muslim 500 publication added that since converting to Islam in 1968, she “has spent the past five decades faithfully learning the Islamic tradition and making its key texts available to the global English-speaking Muslim community, sometimes in collaboration with her husband, Abdalhaqq Bewley, with whom she translated the noble Qur’an.”
Married to Hajj Abdalhaqq Bewley, Ms Bewley has written and translated a number of works on Islam to open up the faith to a wider audience because it had previously only been available in Arabic."
#metaglossia mundus