Dictionnaire peul-français (Mali) - Presses de l'Inalco - Seydou Christiane

Dictionnaire peul-français (Mali)
Christiane Seydou
Les Presses De L'Inalco 19 Novembre 2026
Dictionnaires bilingue français / langue étrangère


"Résumé
L'aboutissement du travail d'une vie de recherche sur la langue peule parlée au Mali.


Répartis sur une dizaine de pays d'Afrique de l'Ouest, dans des îlots de peuplement formant un « archipel », les Peuls partagent une culture commune qui s'appuie d'abord sur l'emploi d'une même langue.
Ce dictionnaire s'intéresse en particulier au Maasinankoore, ou le peul du Massina, région du Mali occupant principalement la courbe du fleuve Niger, qui est selon l'expression consacrée le « nombril du monde peul ». Pour Christiane Seydou, spécialiste du patrimoine littéraire peul du Mali, c'est l'aboutissement du travail d'une vie, compilant le lexique récolté lors d'entretiens sur le terrain, et issu de textes poétiques, d'articles et d'ouvrages spécialisés. Il est enrichi en annexe d'une analyse des motifs des couvertures et tentures, ainsi que des marques de bétail. Ce nouveau dictionnaire bilingue a ainsi vocation à être un instrument pédagogique à la disposition des chercheurs sur le terrain et une base de travail pour l'élaboration d'un dictionnaire pluridialectal.


Directrice de recherche honoraire au CNRS, Christiane Seydou est spécialiste du patrimoine littéraire peul du Mali, agrégée de grammaire et diplômée de peul et de haoussa. Elle a notamment publié des éditions bilingues d'oeuvres littéraires peules : épopée, conte, poésie profane et religieuse."
Dictionnaire peul-français (Mali) - Christiane Seydou - Les Presses De L'Inalco - Grand format https://share.google/nYuUUjD7mfnkqNGxz
https://share.google/s8l6wJCWMRxj7Ie9x
#metaglossia
#metaglossia_mundus

When a Model Writes Uzbek in Three Alphabets – Communications of the ACM

One user writes Uzbek in Cyrillic, another in Latin, a third in Latin with three different apostrophes


"Some users will read an answer in a different language as a sign that the model wasn’t built for them.


By Ivan Lapa
Posted Nov 16 2026


When a team in Tashkent or Almaty ships an AI feature, whether a support bot, menu search, or voice-note parsing, the first debate is almost always about the model: which one understands Uzbek, whether to use a Kazakhstan-built model, whether GPT is good enough. Most failures happen upstream of the model. One user writes Uzbek in Cyrillic, another in Latin, a third in Latin with three different apostrophes, and many messages are partly Russian. A model that looks strong on a benchmark behaves very differently on that stream.


The script problem is about to get harder. Uzbekistan began moving from Cyrillic to Latin script in 1993; the deadline for the change slipped from 2000 to 2005, 2010, and then 2023, and both scripts remain in wide use. On October 21, 2026, Uzbekistan President Shavkat Mirziyoyev signed a law revising the Latin alphabet itself: from April 22, 2027, sh becomes Ş, ch becomes Ç, o‘ becomes Ö, and g‘ becomes Ğ, with no single deadline for the switch. Uzbek text will arrive in three written forms, because old text and typing habits remain. Even today, o‘ is typed as oʻ, o’, o` or o’, since the correct character isn’t on a standard keyboard. A person reads one word; your search index, tokenizer, and string matching see four. Kazakhstan’s own Latin transition still has no timeline, and Kazakh online is mostly Cyrillic, with Russian mixed in mid-sentence.


What do the benchmarks say? KazMMLU, comprised of about 23,000 questions in Kazakh and Russian, offers two lessons. Global frontier models lead by a wide margin: with Kazakh instructions, GPT-4o averages 76.7% accuracy, while two Kazakh-specialized eight-billion-parameter models land around 38–39%. And results depend on the language of the instructions in ways that vary by model: GPT-4o barely moves between English and Kazakh instructions, while in the KazMMLU paper’s tables DeepSeek V3 falls from 76.9% to 45.5%. That is one model in one run, but the practical lesson holds: test with the exact prompt you will ship.


TUMLU is unusual in testing both scripts. On Kazakh, all three tested models performed better in the Cyrillic script, and Latin-script prompts often pushed them into output that resembled Tatar. On Uzbek, the script made little difference (Claude 3.5 Sonnet scored 67.9 in Cyrillic, and 68.6 in Latin), because Uzbek text online is split more evenly. The paper also carries a warning: Uzbek looked most difficult partly because its questions were modeled on university entrance exams. Comparing languages by one number from someone else’s benchmark mostly compares the people who wrote the questions.


Three more gaps matter for products. Code-switching is barely studied: researchers working on Kazakh-Russian translation found no public parallel corpus and built one of 618 sentences. For voice, the Whisper paper found that recognition error roughly halves with every sixteen-fold increase in training data for a language, so strong Russian recognition says nothing about Uzbek. And tokenizers can turn the same text into sequences up to fifteen times longer depending on the language, which moves price, latency, and how much history fits in context. Count tokens per message for each language and script in your own traffic before you sign off on unit economics.


So normalize before the model. Detect language and script per message, not per user profile: the same person may type Latin on a phone in the morning and Cyrillic on a laptop at night. Map all apostrophe variants to one canonical character in a copy used for search and matching, and keep the original. Index both scripts, using transliteration tables built with a native speaker and tested on real dish, street, and product names. Don’t convert Kazakh to Latin before the model for consistency’s sake, since TUMLU suggests Cyrillic works better. Reply in the script the user wrote in. And store the alphabet version as a data field, so that a search conducted after April 2027 finds a word in any spelling without relabeling the archive.


Then build your own evaluation set from real requests, split into strata: Uzbek Cyrillic, Uzbek Latin (soon in both orthographies), Kazakh, Russian, mixed messages, and voice. Score each stratum separately; if most traffic is Russian, an overall average will hide a failure on Uzbek Cyrillic. Use native-speaker labelers from different generations, since people schooled before and after the switch to Latin write differently, and trust an LLM judge only on strata where you have checked it against human ratings.


Finally, treat fallback as a product decision. Some users will accept a Russian reply to an Uzbek question; others will read it as a sign that the product wasn’t built for them. When confidence is low, ask a clarifying question in the user’s language or offer buttons. In payments, safety, or account blocks, route to a person who speaks the language. Offer Russian as a choice rather than switching silently, and log every fallback with its language and script, because a month of logs shows where the model fails systematically.


Where to start? Pull a week of real messages and label them by language and script. It takes a couple of days and shows what traffic you are actually serving. You can swap a model in a week; your normalization layer and evaluation set will outlast several model changes and the move to Uzbekistan’s new Latin alphabet.


Disclosure: The views expressed are my own and do not necessarily reflect those of my employer.


Ivan Lapa
Ivan Lapa is a product manager at Uzum Tezkor in Tashkent, Uzbekistan. He previously worked on courier products at Yandex Pro."
When a Model Writes Uzbek in Three Alphabets – Communications of the ACM https://cacm.acm.org/blogcacm/when-a-model-writes-uzbek-in-three-alphabets/
#metaglossia
#metaglossia_mundus

Lu Xun’s Overlooked Translations That Shaped Modern Chinese Literature

Lu Xun’s Overlooked Translations That Shaped Modern Chinese Literature
Long before becoming one of China’s most celebrated writers, Lu Xun turned to foreign works in search of new ways to write, think, and understand the world. His work would help transform not only Chinese writing but also the language itself.
By Emily Xueni Jin
Nov 18, 20267-min read #translation#literature


Emily Xueni Jin
Researcher
Emily Xueni Jin is a scholar of Chinese literature, media, and techno-humanities with a Ph.D. from Yale University. She is also a sci-fi and fantasy translator.


In 1903, 22-year-old Lu Xun, then a student in Japan, wandered into a bookstore in Jimbocho, Tokyo’s famed book district, where his attention was caught by a science fiction novel titled “A Journey to the Moon in Ninety-Seven Hours and Twenty Minutes.” Translated into Japanese by Inoue Tsutomu and believed to have been written by an American author, the novel appealed to Lu Xun’s own ambition to popularize scientific knowledge in China, where modern science education was still in its infancy. Enthusiastic about the genre’s potential, he quickly set about translating it into Chinese.


His version, “Yuejie Lüxing” (“A Journey to the Moon”), published the same year and among the earliest Chinese translations of science fiction, marked the beginning of his prolific career as a translator, long before he became celebrated as the “founding father of modern Chinese literature” for his novels and essays.


Today, we know that Inoue had misidentified the original author. The novel was, in fact, Jules Verne’s “De la Terre à la Lune: trajet direct en 97 heures 20 minutes,” published in 1865 and widely translated into English in the 1870s. Misled by its setting — in post-Civil War America, where members of the Baltimore Gun Club plan to fire a projectile at the moon — Inoue assumed the author was American. Lu Xun, following this assumption, opened his translation with original passages describing America in extravagant detail to provide context for his Chinese readers.


SUBSCRIBE TO OUR NEWSLETTER
Enter your email here
Submit
By submitting, you agree to our Terms Of Use.
*Please enter an email address.


Left: The title page of Lu Xun’s translation of “A Journey to the Moon.” Right: The copyright page of Lu Xun’s translation of “Journey to the Center of the Earth.” From the collection of the Museum of Modern Chinese Literature.
The curious story of this translation and mistranslation of Verne’s novel illuminates how literary forms and ideas traveled across borders at the turn of the 20th century, particularly in East Asia following the First Sino-Japanese War (1894–1895), when waves of Chinese students — including Lu Xun — went to study in Japan.


It was in Tokyo, amid the intellectual riches of Jimbocho, where he spent much of his leisure time, that Lu Xun was drawn to translation. Japan’s Meiji Restoration, which restored imperial rule in 1868, had already spurred a flood of translations from European languages into Japanese. Eager to take advantage of this wealth of material, Lu Xun, by then proficient in Japanese, with a working knowledge of German and some Russian, immersed himself in these texts. On one hand, he translated texts from cultures he perceived as more scientifically and technologically advanced, with the aim of promoting science education among his own people; on the other hand, he gradually developed an interest in writers from marginalized communities and nations, believing their experiences would resonate with China, which had endured decades of foreign invasion and exploitation. Later, recounting his decision to translate the exiled Russian writer and activist Vasily Eroshenko, he made clear that he saw translation not simply as an aesthetic pursuit, but as a means of giving voice to the oppressed.


It is worth remembering that Lu Xun, then an unknown young student named Zhou Shuren, had yet to adopt his most famous pen name that would be enshrined in modern Chinese literature. In 1918, he published “A Madman’s Diary,” a pioneering work of modern Chinese fiction inspired by a short story of the same name by Russian writer Nikolai Gogol. He later wrote the celebrated novella “The True Story of Ah Q,” published in 1921, helping establish vernacular Chinese as a language of modern literature. He is remembered as a pioneer of the New Culture Movement, which championed “Mr. Democracy” and “Mr. Science” just as China transitioned from monarchy to republic.


A view of Uchiyama Shoten in Tokyo’s Jimbocho district, Japan, 2022. Xinhua
But in 1903, a literary career was not yet his ambition. Funded by a Qing government scholarship to study in Japan, he began by taking language courses, translating scientific essays, science fiction, and literary works along the way, before enrolling in medical school in Sendai City, 370 kilometers north of Tokyo, in 1904.


He would soon abandon medicine to pursue his own literary and intellectual revolution, with translation as his chosen medium. Back in Tokyo in 1906, Lu Xun remained enrolled at a German-language school, but rather than attend classes, he devoted his time to translating, writing, and publishing. He shared a small room with his younger brother, Zhou Zuoren (1885–1967), who had joined him in Japan.


The brothers sent their translations to publishers in China. Translation was poorly paid a century ago, just as it often is today. Zhou Zuoren later recalled with resignation: “… manuscript fees were truly meager. Typically, the translation of Western-language texts only earned two yuan per 1,000 characters — and this was a net amount, with all punctuation and black spaces excluded from the calculation.”


Nevertheless, the brothers pursued their work with passion, publishing “Collection of Stories from Abroad” in . Reflecting their intellectual ambitions, the collection featured revolutionary literature from “peripheral” nations whose experiences echoed China’s own. They hoped such works could change minds and contribute to social progress. Funded by Jiang Yizhi, a banker from China’s eastern Zhejiang province, the collection sold poorly, with just 21 copies of the first volume and 20 of the second sold in six months, according to Lu Xun’s preface to a later edition.


Left: The title page of “Collection of Stories from Abroad.” Right: The preface from the manuscript of “Collection of Stories from Abroad,” in which Lu Xun wrote, “New masters of literary art from different lands can hence begin their entry onto the soil of China.” Photographed by Emily Xueni Jin at the Lu Xun Museum.
Like translators today, those of the past could find their intellectual lives both poor and rich: financial hardship drove them to scrimp and work harder, while literature’s riches tempted them to procrastinate. We might imagine these “revolutio.nary figures” readily sacrificing leisure for national salvation, but their personal accounts tell a more relatable, even comforting, story.


After Lu Xun returned to Tokyo from Sendai, the brothers moved to the wider Kanda-Jimbocho area. After moving a few times, they rented the former residence of renowned Japanese novelist Natsume Sōseki with three friends and named their share-house Wushe, “the house of five.” As admirers of Natsume, they even shopped at the same grocery stores he had frequented and followed his recommended reading list. Personal accounts suggest that Lu Xun spent much of his stipend on books, economizing on food and other daily necessities as a result, which also helps explain why he shared a room with Zhou Zuoren.


Tightening their belts for books while working on their envisioned projects, Zhou Zuoren described a conflict with Lu Xun when they shared a room: “He kept urging me to translate, but I brushed him off. One day, he suddenly flew into a rage, raised his fist, and beat me on the head several times.” As a translator who habitually procrastinates, I can’t bear to imagine what life would be like if I shared a room with my editor.


Translation was more than just work for the Zhou brothers. They lived a “translated” way of life. Kanda-Jimbocho, where they lived, served as a “universal library” and remains Tokyo’s most celebrated book district. Much like Chinese students abroad today who frequent bookstores such as Strand Bookstore, Daunt Books, Shakespeare and Company, or Kinokuniya, the brothers visited Jimbocho’s bookshops almost daily.


The brothers devoured texts across languages and genres. They kept detailed records of their favorite bookstores, even rating the quality of their books and service. Their daily lives involved constant movement between languages as they learned new ones and absorbed political, social, and literary theories, whether in their original languages or in translation.


Left: Photographs presented by Lu Xun to Xu Shouchang. The reverse bears Xu’s inscription: “Lu Xun’s 24-year-old portrait in Tokyo in 1904.” Photographed by Emily Xueni Jin at the Lu Xun Museum. Right: The former Tokyo residence of Natsume Sōseki, later rented by Lu Xun and four friends as Wushe (“House of Five”). From the collection of the Lu Xun Museum.
This broad and intensive engagement transformed their intellectual horizons, offering access to ideas from around the world. As Zhou Zuoren later recalled, Lu Xun would read and write until dawn, sleep until late morning, get up to smoke, and then repeat the cycle.


Later, figures such as Zhou Enlai, the first premier of the People’s Republic of China; Guo Moruo, a writer, poet, and communist leader; Tian Han, the playwright who wrote the lyrics to China’s national anthem; and Feng Zikai, a pioneer of modern Chinese comics, would undergo similar intellectual journeys shaped by translation while studying in Tokyo. They, too, explored different paths toward China’s modernization through their engagement with translated texts, ideas, and ways of life.


In a 1931 letter to fellow writer and revolutionary Qu Qiubai, Lu Xun argued that “translations not only introduce new things but also new ways of expression.” By incorporating unfamiliar syntax into modern Chinese through translation, whether “ancient, dialectal or foreign,” he believed “we can later make it our own.”


His goal was to shape modern Chinese into a truly global language through translation, one flexible enough to absorb diverse forms of expression and capable of conveying ideas from around the world. Arguably, in his imagination, modern Chinese was a language born of translation, one that proudly celebrates its linguistic inheritances. Tian Han and Feng Zikai shared similar convictions, believing that drama and comics — beyond translation in the narrow linguistic sense — embodied cosmopolitan ideals and egalitarian possibilities more vividly than other forms.


To fully appreciate Lu Xun’s significance in modern Chinese literature and language, we cannot examine his fiction alone. In many ways, his identity as a translator — both in the literal sense and in the broader, metaphorical sense of absorbing influences from other cultures and passing them on — was fundamental to his intellectual and literary work.


At a time when cultural exchange has become increasingly complex and contentious, Lu Xun’s story as a translator offers a fresh perspective on what it means to be truly global.


(Header image: Lu Xun (left) with Kaizōsha"
https://www.sixthtone.com/news/1019257
#metaglossia
#metaglossia_mundus