[unable to retrieve full-text content]
A U.S. Assistant Treasury Secretary's Passion for the Chassidic Masters - Lawyer, government official and China ... Chabad.orgWednesday, January 31, 2024
First Mental Health Metaphor Dictionary to raise awareness of disorders - Medical Xpress - Dictionary
Metaphors are not just literary devices for writers to embellish their texts. They are linguistic tools used in everyday life, in most cases with the aim of better understanding and conveying the reality of the world around us.
Researchers at the Universitat Oberta de Catalunya (UOC) have now developed the first Mental Health Metaphor Dictionary, a pioneering repository that brings together and exemplifies the most important conceptual metaphors used in Spanish by people with serious mental illnesses, mainly schizophrenia, bipolar disorder, depression and obsessive-compulsive disorder.
The repository is based on first-person accounts of what it is like to live with a particular mental disorder and can be useful for mental health communicators and professionals, relatives of people diagnosed with one of these illnesses and even for the people with these disorders themselves. The dictionary is presented in an article published in Digital Scholarship in the Humanities.
According to Marta Coll-Florit and Salvador Climent Roca, researchers from the Linguistic Applications Interuniversity Research Group (GRIAL) of the Faculty of Arts and Humanities, who coordinated the development and creation of the repository, "This is a tool to raise visibility and awareness in society about the suffering of people with mental illnesses. We believe that this dictionary can help us all to reflect on the way we talk about mental health and to realize the power of our words."
Metaphors are not neutral
The dictionary is one of the results of the MOMENT project, which seeks to identify the metaphors used in the field of mental health and the underlying conceptualizations. The researchers said, "The metaphors we use are not neutral, but have the power to highlight certain aspects of reality while potentially obscuring others. They're likely to reveal latent ideas that don't come out explicitly in our discourse, but may be filtered through figurative language.
"For example, it's not the same to say 'you have to fight your illness' as it is to say 'you have to live with your illness': The first metaphor emphasizes the struggle between the person and the disorder, whereas the second emphasizes the person's acceptance of their situation."
One of the project's main research conclusions is that this type of metaphor can have beneficial or harmful uses in public discourse and in the discourse of those affected and those who interact with them. "Beneficial uses are those that convey empowerment, control or positive emotions; in other cases, they look to present a problem by separating the negative aspects from the situation as a whole," they explained.
Metaphors that serve these purposes are called "empowering metaphors" and are recommended for use "in public discourse and in relationships with people diagnosed with mental disorders." The researchers added that "their use should be encouraged by the people with these disorders themselves in order to avoid pejorative views of their situation."
A window into the feelings of people with mental disorders
The dictionary is organized both alphabetically and thematically into three broad areas: metaphors of living with a mental disorder, metaphors of communication and social context, and metaphors of medicine and professional practice.
All the metaphors are grouped around different key concepts and come with several examples. As the researchers explained, by systematizing and exemplifying the metaphors used by these people, we can gain a deeper insight into what they "really think and experience." It is also a way for them to feel "more understood and less alone, realizing that their feelings and experiences are shared by more people."
The advantages of blogs and social media
One of the key features of the repository is that all the metaphors are taken from texts posted in Spanish on blogs or on X (formerly Twitter). These communication channels have an important advantage over other written media.
"The authors' words are not filtered by an external interviewer, but come from a genuine and spontaneous willingness to share a lived experience on social media. In addition, individuals can use the relative anonymity of the internet to reveal things they would not, for example, discuss in a face-to-face research setting. For this reason, the range of metaphors found is much wider than in previous similar studies."
This approach has been useful in collecting the many metaphors in the repository that criticize the medical profession or highlight the suffering caused by social stigma and discrimination, showing "how patients seek greater empathy and understanding of their suffering from both medical staff and the wider community." In this regard, the researchers stressed that the dictionary can be valuable in promoting "more respectful discourse" on mental health by public institutions and the press.
A tool for detecting psychopathology
Finally, the Mental Health Metaphor Dictionary can be used as a gateway to detecting psychopathology. "Knowing which conceptual metaphors are most commonly used to express mental distress could help families or people close to those affected to identify possible disorders," the researchers explained.
Likewise, although it cannot be used directly as a diagnostic tool, it could be used to "identify which stage of the disorder the patient is in, according to the type of metaphors used, or to analyze whether therapy has been successful."
A pioneering initiative
This collection of metaphors is the first of its kind. While there are a number of domain-independent repositories of conceptual metaphors based on English texts, there are very few domain-specific repositories in other languages.
In fact, in the review, they found only two repositories focused on specific subject areas: cognition and health. "Although these subject repositories are potentially the most useful for society, they are the rarest," they concluded.
More information: Marta Coll-Florit et al, Metaphor repositories: the case of the mental health metaphor dictionary, Digital Scholarship in the Humanities (2023). DOI: 10.1093/llc/fqad058
Provided by Universitat Oberta de Catalunya (UOC)
Citation: First Mental Health Metaphor Dictionary to raise awareness of disorders (2024, January 31) retrieved 31 January 2024 from https://ift.tt/SITFNKa
This document is subject to copyright. Apart from any fair dealing for the purpose of private study or research, no part may be reproduced without the written permission. The content is provided for information purposes only.
How Large Language Models Fare Against 'Classic' Machine Translation Challenges - Slator - Translation
In a January 17, 2024 paper, a group of researchers from the University of Macau, University College London (UCL), and Tencent AI Lab explored the performance of large language models (LLMs) against “classic” machine translation (MT) challenges.
The six MT challenges, originally proposed by Philipp Koehn and Rebecca Knowles in 2017, include domain mismatch, amount of parallel data, rare word prediction, translation of long sentences, attention model as word alignment, and sub-optimal beam search.
For their experiments, the researchers used the Llama2-7b model, focusing on the German-to-English language pair. They explained that “English and German are high-resource languages in the Llama2 pretraining data, which ensures the model’s proficiency in these two languages.”
They found that LLMs reduce dependence on parallel data during pretraining for major languages and they improve the translation of long sentences and entire documents. Yet, challenges like domain mismatch and rare word prediction persist. Unlike neural MT models, LLMs face new challenges: translation of low-resource languages and human-aligned evaluation.
Document-Level
Specifically, the researchers found that LLMs mitigate reliance on bilingual data during pretraining for high-resource languages, with even a small amount of parallel data boosting translation performance. Surprisingly, an increased abundance of parallel data yields only marginal improvement and, in some cases, a decline in LLM translation system performance, challenging the common belief that more parallel data enhances translation quality. The researchers recommended supervised fine-tuning as a more advantageous approach for leveraging additional parallel data compared to continued pretraining.
The research community should “consider how to efficiently utilize parallel data for the enhancement of LLM translation systems, thereby offering a potential direction for future studies to optimize bilingual knowledge in the pursuit of improved MT performance using LLMs,” according to the researchers.
Another addressed challenge was the translation of long sentences, a significant hurdle for MT systems. LLMs demonstrated an ability to tackle this challenge effectively excelling in translating sentences with fewer than 80 words and consistently performing well at the document level with approximately 500 words.
“LLMs excel in translating extended sentences and entire documents, underscoring their effectiveness as a promising solution for addressing challenges associated with long-sentence and document-level translation tasks,” they said.
Unresolved Challenges
The researchers explored whether the rich knowledge of LLMs could address domain mismatch in translation tasks. While LLMs showed robust performance in in-domain translation tasks, their progress in out-of-domain tasks was modest, encountering challenges like terminology mismatch, style discrepancies, and hallucinations.
Predicting rare words in the realm of LLMs remains another significant challenge, leading to omissions in translations. The researchers underscored the persistent and unresolved nature of this issue, emphasizing its significance in the field.
Mixed Results
Word alignment, involving the identification of word pairs with similar semantic information in a given translation pair, was also explored. The researchers tested the feasibility of extracting word alignment from LLM attention weights, revealing that it was not a viable option. Despite this, the process provided valuable insights into model interpretability, they said.
In the context of inference, two major issues are inference strategies — including beam search and sampling — and inference efficiency due to the abnormal size of LLMs, as the researchers explained. They first tested the performance difference of beam search and sampling and they found that beam search is not necessarily suboptimal in LLMs.
In terms of inference efficiency, they found that LLMs require an average of 30 seconds compared to the 0.3 seconds of MT models, raising concerns about real-time deployment in scenarios requiring fast translation. “The longer inference time of LLMs may impede their real-time deployment in scenarios where fast translation is required,” they said.
New Challenges
Besides these six “classic” MT challenges, they identified two new challenges within the realm of LLMs. One pertains to the translation quality for language pairs inadequately represented during the pretraining stage and the other involves evaluating translation quality.
The researchers found that translation performance is significantly affected by the available resources for each language, emphasizing the need for a diverse and balanced dataset during the pretraining of LLMs to ensure equitable performance across languages.
Evaluation issues have also come to the forefront. They tested the quality of LLMs using both automatic — BLEU and COMET — and human evaluation metrics and found a moderate negative correlation between them. This emphasizes the importance of combining both evaluation methods and indicates that current metrics may not fully capture the nuances appreciated by human evaluators.
According to the researchers, this calls for further research to develop and refine evaluation methods aligned with human preferences, especially as language models become more complex and capable. “This human-centered approach to evaluation will be crucial in ensuring that our translation models are not only technically proficient but also practically useful and acceptable to end users.” they said.
Finally, the researchers called for future research to focus on refining evaluation methods and testing approaches on more advanced models.
Authors: Jianhui Pang, Fanghua Ye, Longyue Wang, Dian Yu, Derek F. Wong, Shuming Shi, and Zhaopeng Tu.
Tuesday, January 30, 2024
Paris to use AI translation app for Olympics visitors - Yahoo News - Translation
STORY: (English) "How can I go to the Stadium of France?"
(Arabic): "How can I get to Stade de France?"
(Korean): "How do I get to the Olympic opening ceremony?"
As Paris welcomes the world for the Olympics this summer, the city’s public transport system, known as RATP will be using artificial intelligence to help thousands of international visitors navigate through the capital.
The handheld Tradivia device can translate between French and 16 different languages including Mandarin, Arabic and Korean, with text appearing on a screen as well as being read out loud.
The RATP will provide more than 3,000 agents with this device, ready to assist all international visitors.
RATP representative Gregoire de Lasteyrie:
"The goal for us in Ile-de-France Mobilite is for them to travel in the best possible conditions, and therefore, being able to speak to them in as many languages as possible and helping them find their way in Paris is extremely important."
Metro workers, like Raphael Gassette, say the device gives them more confidence.
“We no longer have this fear of thinking, 'We're not going to understand each other.' Here, we know straight away, with regard to the languages, to press straight on 'Hindi,' and immediately have clear, more precise information, and we can be sure that when the visitor leaves, they are satisfied."
The service will remain in Paris after the Games, which will be held from July 26 to Aug. 11.
American Institutes for Research Honored by Anthem Awards for Knowledge Translation Expertise - Yahoo Finance - Translation
Anthem Awards 2024 Winner
Arlington, Va., Jan. 30, 2024 (GLOBE NEWSWIRE) -- The American Institutes for Research (AIR) has won two Anthem Awards for its work operating the Model System Knowledge Translation Center (MSKTC), which translates health information into easy-to-understand language and formats for people living with spinal cord injury (SCI), traumatic brain injury (TBI), and burn injury and their families and caregivers. The Anthem Awards, presented by the International Academy of Digital Arts and Sciences, recognizes innovation in the work of mission-driven organizations committed to amplifying social impact causes that spark global change.
The MSKTC was recognized in the community engagement and public service categories for its innovative approach to engaging Spanish speaking audiences through the translation and dissemination of pertinent health information to meet the needs of individuals with traumatic injuries. MSKTC’s Spanish language resources earned gold recognition in the nonprofit sector for best-in-class Public Service health project and a silver honor in the category of best-in-class Community Outreach project.
“We are honored that our work to support people living with SCI, TBI, and burn injuries is resonating across communities and our broader society across the globe,” said Xinsheng “Cindy” Cai, principal researcher at AIR who directs the MSKTC. “We will continue our commitment to overcoming language barriers and ensure the latest research findings are being used in health care decision-making.”
As a federally funded national center, the MSKTC works closely with medical professionals from the Model System centers, who conduct innovative and high-quality research and provide multidisciplinary rehabilitation care to meet the information needs of individuals living with SCI, TBI, and burn injury by identifying health information needs, summarizing research, and developing and disseminating information resources. Both the MSKTC and Model System centers are funded by the National Institute on Disability, Independent Living, and Rehabilitation Research (NIDILRR), Administration for Community Living, U.S. Department of Health and Human Services.
Since 2021, the suite of digital resources on the MSKTC website has received more than 7.1 million pageviews from 4.1 million visitors globally, reaching Spanish-speaking countries, including Mexico, Spain, Columbia, and Argentina. In total, the project has reached 240 countries, reflecting AIR’s strong commitment to evidence-based learning and increasing equitable access to education and health services.
As a winner of the Anthem Awards, AIR is in the company of other distinguished organizations, including AARP, the CDC Foundation, the Johns Hopkins Bloomberg School of Public Health, and others.
Launched in 2021 by The Webby Awards, the Anthem Awards honors the purpose and mission-driven work of people, companies and organizations worldwide. Social impact projects are evaluated across seven core causes: Diversity, Equity, and Inclusion; Education, Art, and Culture; Health; Human and Civil Rights; Humanitarian Action and Services; Responsible Technology; and Sustainability, Environment, and Climate. This year, more than 2,000 entries spanning 30 countries were submitted.
About AIR
Established in 1946, the American Institutes for Research (AIR) is a nonpartisan, not-for-profit institution that conducts behavioral and social science research and delivers technical assistance both domestically and internationally in the areas of health, education, and the workforce. AIR's work is driven by its mission to generate and use rigorous evidence that contributes to a better, more equitable world. With headquarters in Arlington, Virginia, AIR has offices across the U.S. and abroad. For more information, visit www.air.org.
Attachment
CONTACT: Dana Tofig American Institutes for Research 202-403-6347 dtofig@air.org
Jimmy Failla's new book 'Cancel Culture Dictionary' puts spotlight on the outrage era plaguing society - Fox News - Dictionary
Jimmy Failla’s book "Cancel Culture Dictionary" hit stores on Tuesday, putting a spotlight on the outrage era plaguing society.
"Cancel Culture is the only movement where the biggest winners are a bunch of losers," Failla told Fox News Digital.
"It's populated by people who spend all day on social media looking for something to get offended by so they can leverage the world's outrage into their ‘likes,’" he continued. "As the book shows, nothing has been improved in the outrage era -- crime is higher, test scores are lower, and we're all a lot fatter despite what the Instagram filters show you."
JIMMY FAILLA'S 'CANCEL CULTURE DICTIONARY' AIMS TO HELP AMERICANS WIN THE WAR ON FUN
"Cancel Culture Dictionary: An A to Z Guide to Winning the War On Fun," is the latest offering from Fox News Books. Failla, a former New York City taxi driver, said the purpose of the project was to simply show how cancel culture has "broken our compass."
He also believes cancel culture has gotten to the point where things aren't being canceled "in the name of progress," but rather for power or personal gain. He cited everything from backlash to comedian Dave Chappelle, the vanishing of syrup icon Aunt Jemima and a school declaring that Abraham Lincoln didn’t prove that Black lives matter to him as some of the most egregious examples of cancel culture gone wrong.
CANCEL CULTURE IS GETTING CANCELED AND IT'S ABOUT TIME
Failla feels that people don’t know the difference between a joke and a hate crime these days, but hopes the latest offering from Fox News Books can help right the ship.
"This book is my attempt to get society back on track. And yes, I'm aware of just how bad things have gotten if a former New York City cab driver who plays video games in his 40's is now the voice of reason," Failla said.
"In short, this book may not save the world," he continued. "But if you like reading at a third-grade level you'll still be glad you bought it."
ORDER ‘CANCEL CULTURE DICTIONARY’ HERE
Failla previously called the book "a step-by-step guide to how everybody can live their life in a way, you know, that will really recalibrate society."
"It's not a call to arms. It's a call to chill the f--k out," Failla said.
"Cancel Culture Dictionary" is available now.
CLICK HERE TO GET THE FOX NEWS APP
Brian Flood is a media editor/reporter for FOX News Digital. Story tips can be sent to brian.flood@fox.com and on Twitter: @briansflood.
Monday, January 29, 2024
A New Translation Unveils The Dark Side Of The Travels Of Marco Polo - Worldcrunch - Translation
ROME — When I am in charge of a translation, the first question I ask myself is not about the text itself, but how to approach the work.
For example, when I embarked on translating Travels in the Congo, the travel diary by French author André Gide, I decided to experiment with an idea I had had for some time: to try and reproduce the original French version in Italian. The book's language was sometimes contracted and fragmentary, sometimes stretched out into lyrical slashes or even soaring in invective. So my mission was difficult: I needed not only to translate it in the most faithful way possible, but in the most corresponding way too, almost as if I was copying it, creating an imitation rather than a translation.
This experiment — of course it will be the readers who will assess how successful it was — came to me out of a growing dissatisfaction with the fact that good, even excellent translations may completely lose the syntactic, grammatical relationship to the original version.
I love languages both for their lexicon than for their structure, for the way they order words in a precise sequence. The notion that a translated text should not "feel" translated, but on the contrary it should turn out as if it had been written in the target language in the first place, is useful for those who translate a lot and work with books of wide circulation. When dealing with refined and rare texts, in my opinion, it is necessary to aim for a more radical approach.
Marco Polo's story
And here we come to Marco Polo's Il Milione (The Million), better known in English as The Travels of Marco Polo, which I have recently translated into modern Italian, a language that even fifteen-year-olds can read.
Il Milione (according to the most credited hypothesis, the title derives from Emilione, the nickname of the Polo family) was penned by Rustichello da Pisa, a modest 13th-century novelist who wrote in the Franco-Venetian language, at the dictation of Marco Polo the explorer.
They both found themselves prisoners of the Genoese, and they both had a desire to put to good use that time of enforced confinement and inactivity. Marco Polo is for all intents and purposes the source and creator, but the language and the hand are Rustichello's.
Beauty and money, discovery and interest, travel and enrichment, pomp and war.
This happened precisely at the end of the 13th century. The original manuscript was soon lost due to the many copies that were made: a case of disintegration due to too much success, fatal at the time of technical non-reproducibility. There are therefore several copies, threaded (how faithfully?) from the original text, and a translator, in the first place, must select one.
My choice fell on the codex that, according to scholars, is the closest to the original, the manuscript named Fr. 1116 in the National Library in Paris. Fr. 1116 has specificities that are not at all original: breaks, interruptions, abrupt suspensions, even lacunae and omissions that, far from making it shoddy or shapeless, characterize it and make it singularly modern.
An illustration of Kublai Khan by Évrard d'Espinques for "The Travels Of Marco Polo."
Wikimedia
Earthy language to unveil
It is so modern that the Tuscan version of the fourteenth century, by a translator who worked on a codex very similar to the manuscript preserved in Paris, perceived an uneasiness in it, and considered rounding off the ending with a spurious epilogue that guaranteed a circularity and a happy ending. That spurious ending, which in my translation I have obviously taken care not to repeat, was widely considered part of the original story.
Il Milione is a text that is indeed marvelous (in the literal sense of the word) and that justifies the exhilarating and smooth rereadings or transpositions that insist on exoticism, on the titanic figure of Kublai Khan and the astonishing imperial palace of Shangdu, on flora and fauna seemingly out of the dawn of time. But at the same time it is also a book shot through with a pragmatism, a brutality that is absolutely concrete, in which pages and pages are devoted to trade, money, and war. Taut, concise pages, in which every word counts, for example in the detailed description of paper money in use by the Mongols.
As this harsh, ironclad side of Il Milione came to light, my translation took on burnished tones, moving as far apart from the other available modern Italian translations, which willingly yield to the one-sided image of Il Milione as a soft, precious silk cloth embroidered with legends.
And while the beauty and finesse of the artifacts are omnipresent in the text, the truth is that the original Marco Polo never tires of mentioning these objects' economic value, price, and the lavish earnings they guarantee. Beauty and money, discovery and interest, travel and enrichment, pomp and war.
And perhaps it was providential that Marco Polo, in his Genoese captivity, met not a sublime poet, but the earthy Rustichello. There were however legions of translators who provided embellishing, sugarcoating, and smoothing, which I tried to stop at the gates of the citadel.
From Your Site Articles
Related Articles Around the Web
From Japan to the world: How to translate a game - Japan Today - Translation
Behind the global success of Japanese video games lies a delicate task: appealing to overseas players whose expectations on issues such as sexism are increasingly influencing the content of major titles.
With the majority of sales for big games now outside Japan, everything from slang words to characters' costumes must be carefully considered for a global audience.
It is a complex process that has come a long way since the "Wild West" of the 1980s and 90s, one high-profile "localization" team told AFP.
"There were no rules, no 'industry standards', and the quality of localisation could vary greatly from one title to the next," said the SEGA of America team who worked on "Like a Dragon: Infinite Wealth" -- the latest title in the hit "Yakuza" series, which was released on Friday.
Back then, translators faced constraints including too-small text boxes, and sometimes game developers did the job themselves in less-than-perfect English.
It also meant that many games from the era, especially dialogue-heavy ones, never made it out of Japan.
"Thankfully, the industry -- and perhaps more importantly consumers -- have changed a lot since those days, and we are now able to be more faithful to the cultural and emotional content of Japanese games than ever," the SEGA team said.
Localization is now integral to the design process, with international gamers in mind from the start.
One key example is "how Japanese game developers dress their heroines" as the #MeToo movement changes mindsets, said Franck Genty, senior localization manager at Japanese game giant Bandai Namco.
"We tell them that the cleavage is a bit too exposed, or the skirt is a bit too short," he told AFP. "Before, they weren't very flexible, but they've become more proactive on such subjects."
The puzzle of game localization affected the 1980 arcade sensation "Pac-Man", with the direct translation "Puck Man" deemed too risky because it could be vandalised.
Some top-selling games including Mario, Final Fantasy and Pokemon involve fantasy worlds that are not overtly Japanese, offering some flexibility for their adaptation.
But the task becomes trickier for series such as "Yakuza", which are set in real-life locations and use slang from Japan's underworld. Getting it right is important: around 70 percent of revenue from recent titles in the "Yakuza" series is from overseas.
But in recent years, booming interest in manga comics, anime cartoons and wider Japanese culture has made the job easier. "People know what ramen is now... we don't need to say 'noodles' any more," Genty said.
His team at the European headquarters of Bandai Namco has adapted games including the "Tekken" fighting series and the smash-hit role-playing game "Elden Ring" into a dozen languages.
The job is as much a cultural challenge as a linguistic one, said Pierre Froget, localization project manager at Bandai. "The player, whichever country they're from, should understand and feel the same thing as someone playing in the original language," he said.
A better understanding of Japanese culture among players means adaptations can be more subtle -- the "Yakuza" series is now called "Like a Dragon", closer to the original Japanese.
LGBTQ caricatures and sexist cliches have also been axed.
"Many representations which were normal in Japan in the first 'Like a Dragon' games are no longer acceptable today," Masayoshi Yokoyama, the series' executive producer, told AFP. "We ask our teams in the United States and Europe to read the game's script, and they tell us if they see things that wouldn't be acceptable in their country."
Changes often focus on "alcohol, politics or religion", Froget said, while cultural reference points also differ.
"When there are people dressed in black boots and big leather coats, in Europe that could bring to mind a Nazi uniform," he said.
With global release dates now the norm, these decisions must be made under tighter deadlines than before.
And despite improved communication between developers and localization teams, challenges remain -- especially when translating a game into languages other than English.
"Efforts have been made to understand the needs of the English-speaking world," Froget said.
But for German, which has longer sentences and other linguistic quirks, localization is sometimes "seen as an extra difficulty" by design teams.
Even so, Froget believes in his mission: "To create connections to Japanese culture and help Europeans discover its depth, while respecting both the game and the player."
© 2024 AFPWe tested Galaxy S24′s real-time AI translation: it works, sort of - 조선일보 - Translation
[unable to retrieve full-text content]
We tested Galaxy S24′s real-time AI translation: it works, sort of 조선일보Sunday, January 28, 2024
A New Translation Unveils The Dark Side Of The Travels Of Marco Polo - Worldcrunch - Translation
ROME — When I am in charge of a translation, the first question I ask myself is not about the text itself, but how to approach the work.
For example, when I embarked on translating Travels in the Congo, the travel diary by French author André Gide, I decided to experiment with an idea I had had for some time: to try and reproduce the original French version in Italian. The book's language was sometimes contracted and fragmentary, sometimes stretched out into lyrical slashes or even soaring in invective. So my mission was difficult: I needed not only to translate it in the most faithful way possible, but in the most corresponding way too, almost as if I was copying it, creating an imitation rather than a translation.
This experiment — of course it will be the readers who will assess how successful it was — came to me out of a growing dissatisfaction with the fact that good, even excellent translations may completely lose the syntactic, grammatical relationship to the original version.
I love languages both for their lexicon than for their structure, for the way they order words in a precise sequence. The notion that a translated text should not "feel" translated, but on the contrary it should turn out as if it had been written in the target language in the first place, is useful for those who translate a lot and work with books of wide circulation. When dealing with refined and rare texts, in my opinion, it is necessary to aim for a more radical approach.
Marco Polo's story
And here we come to Marco Polo's Il Milione (The Million), better known in English as The Travels of Marco Polo, which I have recently translated into modern Italian, a language that even fifteen-year-olds can read.
Il Milione (according to the most credited hypothesis, the title derives from Emilione, the nickname of the Polo family) was penned by Rustichello da Pisa, a modest 13th-century novelist who wrote in the Franco-Venetian language, at the dictation of Marco Polo the explorer.
They both found themselves prisoners of the Genoese, and they both had a desire to put to good use that time of enforced confinement and inactivity. Marco Polo is for all intents and purposes the source and creator, but the language and the hand are Rustichello's.
Beauty and money, discovery and interest, travel and enrichment, pomp and war.
This happened precisely at the end of the 13th century. The original manuscript was soon lost due to the many copies that were made: a case of disintegration due to too much success, fatal at the time of technical non-reproducibility. There are therefore several copies, threaded (how faithfully?) from the original text, and a translator, in the first place, must select one.
My choice fell on the codex that, according to scholars, is the closest to the original, the manuscript named Fr. 1116 in the National Library in Paris. Fr. 1116 has specificities that are not at all original: breaks, interruptions, abrupt suspensions, even lacunae and omissions that, far from making it shoddy or shapeless, characterize it and make it singularly modern.
An illustration of Kublai Khan by Évrard d'Espinques for "The Travels Of Marco Polo."
Wikimedia
Earthy language to unveil
It is so modern that the Tuscan version of the fourteenth century, by a translator who worked on a codex very similar to the manuscript preserved in Paris, perceived an uneasiness in it, and considered rounding off the ending with a spurious epilogue that guaranteed a circularity and a happy ending. That spurious ending, which in my translation I have obviously taken care not to repeat, was widely considered part of the original story.
Il Milione is a text that is indeed marvelous (in the literal sense of the word) and that justifies the exhilarating and smooth rereadings or transpositions that insist on exoticism, on the titanic figure of Kublai Khan and the astonishing imperial palace of Shangdu, on flora and fauna seemingly out of the dawn of time. But at the same time it is also a book shot through with a pragmatism, a brutality that is absolutely concrete, in which pages and pages are devoted to trade, money, and war. Taut, concise pages, in which every word counts, for example in the detailed description of paper money in use by the Mongols.
As this harsh, ironclad side of Il Milione came to light, my translation took on burnished tones, moving as far apart from the other available modern Italian translations, which willingly yield to the one-sided image of Il Milione as a soft, precious silk cloth embroidered with legends.
And while the beauty and finesse of the artifacts are omnipresent in the text, the truth is that the original Marco Polo never tires of mentioning these objects' economic value, price, and the lavish earnings they guarantee. Beauty and money, discovery and interest, travel and enrichment, pomp and war.
And perhaps it was providential that Marco Polo, in his Genoese captivity, met not a sublime poet, but the earthy Rustichello. There were however legions of translators who provided embellishing, sugarcoating, and smoothing, which I tried to stop at the gates of the citadel.
From Your Site Articles
Related Articles Around the Web
Emily Wilson’s fluent new translation of the Iliad honours the epic poem’s power and beauty - Scroll.in - Translation
A new translation of the Iliad of Homer is cause for general celebration, especially when the translator is Emily Wilson of the University of Pennsylvania. Having turned her hand to translations of other Greek and Latin texts – notably Seneca, Euripides, Oedipus Tyrannos, and Homer’s Odyssey – Wilson has moved on to the Iliad, joining an exclusive club of translators of this great work that includes Richmond Lattimore and Robert Fagles.
This is an excellent publication where some bold decisions have been made to provide a sense of the sound and pace of the original text. As Wilson says in the translator’s note:
I wanted to honour the poem’s oral heritage with a regular and audible rhythm, and with language that would, like the original, invite reading out loud, and come to life in the mouth.
Thus, when reading Wilson’s Iliad one senses something of the chant of Homer’s verse, even through the written word.
A free translation
Wilson’s book is much more than a translation. It contains a detailed introduction to the nature and dating of Homeric verse, the historical and archaeological issues of the Trojan war, the code of honour within which the Homeric heroes operate, and the broader mythical context of the war. The book could be a whole course in itself if you wanted to make it one.
We are reminded, for instance, in a discrete section of the introduction, that the Iliad describes the destruction of Troy and the fate of its women, raped and abused by the conquering Greeks. Wilson writes that the “silencing, rape, subjugation, kidnapping and enslavement of women in war are essential instruments for the construction of male honour”.
The more one engages with the Iliad, the more one sees that it is not just a poem of immense power and beauty. It cast such a spell over antiquity that poets and artists after Homer spent much of their time engaging with it.
The Roman poet Vergil, for instance, whose epic poem The Aeneid (written about 700 years after the Iliad) was also focused on the Trojan War theme, may have known the Iliad by heart. When we pick up Wilson’s translation we realise what a task that must have been.
One of the first things to note about Wilson’s translation is that it makes no attempt to offer line-by-line equivalence with the Greek text, as Lattimore does in his 1951 translation. Thus, the 24 books of the poem have both the original Greek line numbers and the line numbers of her translation. This means students of Homeric Greek will not find Wilson’s text such an easy point of reference to check up on their translations.
For Wilson, it was liberating to free herself from the same number of lines as the original Greek. “Once I understood that I needed more lines than the original,” she writes, “I realised I could sometimes use lovely long polysyllabic English words, echoing the original’s use of powerfully long, often compound words interspersed with many shorter connectives, verbs and particles.”
Inevitably, the plethora of names, and the Homeric penchant for repetition in a broader sense, caused Wilson plenty of hard thinking, not to mention the matter of the epithets – the formulaic phrases that appear throughout the poem. How would she deal with “swift-footed Achilles” rather than just “Achilles”, or “Phoebus Apollo”, or “rosy-fingered Dawn”?
Some of the most prominent and radical research in Homeric scholarship over the past hundred years or so (after Milman Parry, who established that Homer’s poetry was most likely not the work of a single poet) has involved scholarly analysis of the epithets. Wilson’s response is to vary the use of Homeric repetition as determined by poetic considerations:
Like almost all modern translators, I have sometimes varied the phrasing of certain formulaic phrases, usually for sonic or rhythmical reasons. So, for example, Zeus appears in this translation both as “cloud-gathering Zeus” and as “Zeus who gathers clouds together”. Minor variations of this kind seemed to me in keeping with the poetic techniques of the original poem, in which epithets are often chosen for metrical reasons as much as anything.
Such a statement will inevitably provide reassurance for textual purists. The epiphany of the goddess Athena to the Greek warrior Achilles, in the midst of Achilles’ feud with Agamemnon in Book 1, gives us a sense of how this plays out and shows what a fine translator Wilson is:
But then Athena swooped down from the sky.
She had been sent forth by the white-armed goddess
Hera, who loved both men. Athena stood
behind Achilles, son of Peleus,
and grabbed him by his chestnut hair. She was
invisible to everyone but him.
Achilles, startled, turned and recognized Athena. She had bright, unearthly eyes.
His words flew out.
“Why have you come here daughter
of Zeus, the god who holds the royal aegis?
Was it to see the cruel violence
of Agamemnon, son of Atreus?”
In such a busy scene as this, with seven individuals mentioned, both deities and mortals – Zeus, Athena, Hera, Achilles, Peleus, Agamemnon and Atreus – it might have been tempting to take out some of the names. Nothing like this happens and the translation is a lot richer for it.
Devotees of Book 6 will also note how the text flows with remarkable ease:
When Hector finished speaking, Hecuba
went in the house and shouted to her slaves
to go through town and call the older women,
and then she went inside her fragrant storeroom.
In it, she kept her fine embroidered robes,
Made by the women of Sidonia
Whom godlike Paris Alexander brought
to Troy across the wide back of the sea,
on that same journey when he brought back Helen,
the daughter of the mightiest of fathers.
Book 6 is one of the more poignant books of the poem. The Trojan warrior Hector returns to the city from the fighting and talks with the women in his family: Hecuba, Andromache and Helen. It loses nothing in the Wilson translation. The reader might also note the reference to Paris as “Paris Alexander” – a rather brilliant way of engaging with the fact that both names (i.e. “Paris” and “Alexander”) are used to describe him in the Iliad.
A long love affair
So on the one hand Homeric purists should not be concerned about the disappearance of certain names or traditions from the Wilson translation. But this is not always the case.
The final book of the poem tells of the ransom mission undertaken by the Trojan king Priam and an old attendant to retrieve the body of Hector from Achilles, who has refused to give it over. The presence of Hermes is crucial in this particular book of the Iliad, because he is the god who oversees reciprocity and exchange. He acts as a guide to the two old men.
But more often than not in Book 24 (and elsewhere in Homer), Hermes is called “Argeïphontês” (11 times), rather than “Hermes” (nine times). The name Argeïphontês seems to mean “Slayer of Argos”. It refers to a somewhat obscure narrative set in earlier times, in which Hermes killed a monster called Argos by first putting him to sleep and then striking him. The name Argeïphontês, it seems to me, is important in various ways, and it is something of a pity that it is dropped from the poem – although Wilson does maintain the monster-killing tradition by calling Hermes “the giant-slayer”.
Another surprising passage from Book 24 is Hermes’s arrival at Troy and his encounter with the two old Trojan men, Priam and Idaeus:
He reached the Hellespont and Troy. He touched down in a human guise.
He looked like a young man, a magistrate,
with beard first sprouting, the most handsome age.
The humans drove beside the tomb of Ilus,
then at the river made the mules and horses
halt for a drink. Dark night already covered
the earth. Idaeus looked around and noticed
Hermes right next to them and said to Priam …
This passage, when I read it, seemed to me a strange translation, not the least for the references to a “magistrate” (i.e. a youthful Hermes) and “the humans” who drive past the tomb of Ilus. A “magistrate” in Homer’s Iliad? I don’t think so.
I don’t know what Wilson was thinking at this point, but she is alert to the danger of anachronism, which needs “to be balanced against an equally pressing danger: that archaism or unidiomatic English risks suggesting that the Iliad is more alien and more simplistic in its values than it really is”.
My two quibbles about Book 24 don’t add up to much in the context of this big work. I offer them as something to reflect upon. What is important about Wilson as a translator is that she has an unequivocal love for the text, which dictates almost all that she does:
I first began reading Homer in high school, early in the study of Ancient Greek. I liked the Odyssey, but I loved the Iliad with a passionate devotion. I have now lived with this poem for some 35 years.
We may be thankful for her love for the Iliad, and the longevity of it, and her generosity in offering it up to readers with very different backgrounds.
Chris Mackie is an Emeritus Professor of Classics at La Trobe University.
This article first appeared on The Conversation.
Saturday, January 27, 2024
Florida schools banned the dictionary. (Yes, you read that right.) | The Grammarian - The Philadelphia Inquirer - Dictionary
There are plenty of reasons to be glad Gov. Ron DeSantis gave up running for president.
First on the list: Now he can go back home to Florida and return to the important work of banning dictionaries.
Yes, that’s actually happening.
That’s actually happening.
In response to laws passed last year in Florida, the School District of Escambia County — which includes Pensacola, in the state’s northwest corner — removed more than 1,600 books from school libraries and placed them into storage.
Included in that sequester: Merriam-Webster’s Elementary Dictionary, The American Heritage Children’s Dictionary, and Webster’s Dictionary and Thesaurus for Students.
Also on the list: The Autobiography of Malcolm X, Anne Frank’s The Diary of a Young Girl, Ernest Hemingway’s For Whom the Bell Tolls, and William Golding’s Lord of the Flies.
While Florida has passed plenty of problematic legislation in recent years, the current controversy is a result of the state’s HB 1069, which prescribes what Florida students learn related to sex and gender (ideally: nothing at all) and how they learn it (ideally: while wearing potato sacks and puritan bonnets).
In June, a month after DeSantis signed the bill, the School District of Escambia County decided “to adopt emergency rules necessitated by a danger to health, safety, and welfare.” Pending further review, it removed every book it could think of that might give students … information?
One could give Escambia County credit for comprehensiveness: What adolescent hasn’t endeavored their own sex ed by consulting the dictionary?
As with much terrible legislation, the problem lies with the law’s lack of precision. Also: the fact that it simply makes up what it wants certain words to mean.
That’s not how language works.
The opening lines of the bill define the word sex as “the classification of a person as either female or male based on the organization of the body of such person for a specific reproductive role, as indicated by the person’s sex chromosomes, naturally occurring sex hormones, and internal and external genitalia present at birth.”
No reputable dictionary defines sex that narrowly. Neither the Merriam-Webster nor the Oxford English Dictionary definition is what anyone would consider radical wokeism, but, like any dictionary, both contain multiple definitions that encompass the ways that sex appears in our lexicon.
Lawmakers don’t get to decide words’ definitions. Neither do school boards, failed presidential candidates, or even grammar columnists.
From there, HB 1069 places all sorts of restrictions on how people talk about and reference sexual classification — all of them meant to squeeze discussion of sex into a “traditional” binary. It’s hard to say which is worse: when the law is vague and calls into question “any material … which … depicts or describes sexual content,” or when it’s hyper-specific and mandates such instruction as “teaching the benefits of monogamous heterosexual marriage.”
» READ MORE: Defining ‘woman’ is complicated for everyone, including Supreme Court nominees | The Grammarian
Dictating language use from the top down is never wise. Even if someone were to try to do so, it would be a job for linguists and lexicographers — not the Florida Legislature. But those linguists and lexicographers are smart enough to know not to try.
The slightly good news is that, after a round of review, the district appears to have removed the dictionaries from its list of banned books; presumably, students are once again allowed to look up definitions. No such luck for Anne Frank, Malcolm X, William Golding, or Ernest Hemingway, all of whom are among the hundreds of books still pending approval.
It would do DeSantis some good to read a few of the books on that list. Lucky for him, he should have some free time now.
The Grammarian, otherwise known as Jeffrey Barg, looks at how language, grammar, and punctuation shape our world. Send comments, questions, and syllepsis to jeff@theangrygrammarian.com.
NJ law requires translation and interpretation services for 7 non-English languages - NorthJersey.com - Translation
By the end of 2025, state agencies and departments must have translated documents into the seven most spoken non-English languages in New Jersey.
That's the mandate of a state law passed by the Legislature earlier this month. The bill takes effect immediately, with translations of hundreds of government documents to begin on a rolling basis this year.
Bill S2459 requires that "state government entities would be required to undertake document translations at a rate of five languages in the first year and two in the second year."
The seven most spoken foreign languages in NJ
The seven most spoken non-English languages in the Garden State, using Census data, are Spanish, Filipino/Tagalog, Chinese, Hindi, Korean, Gujarati and Portuguese.
NJ:Gov. Phil Murphy taps Francis K. O'Connor as new NJ DOT commissioner
The legislation was passed by the Senate and Assembly on Jan. 8. Gov. Murphy it into law on Jan. 12, along with two other bills focused on immigrant communities, one providing new work protections for domestic employees and another changing the way the state gathers data on the Asian Americans and Pacific Islanders community.
The translation and interpretation measure was primarily sponsored in the Senate by state Senators Teresa Ruiz, Nellie Pou and Nilsa Cruz-Perez and in the Assembly by outgoing Assembly member Sadaf Jaffer and current Assembly members Ellen Park and Sterley Stanley.
After Murphy signed the bill into law, Senate Majority Leader Ruiz said, in a statement, "This legislation will remove the language barrier faced by so many of our communities by ensuring all state entities are prepared to assist our residents regardless of what language they speak."
Agencies must provide interpretation services
The new law specifies that translations in required languages be completed no later than 23 months from the time it is signed. It also directs government agencies to develop and implement a language access plan. The plan includes an assessment of the interpretation needs of residents with limited English proficiency and a way to provide interpretation services.
The $500,000 in funding needed to implement the new law is in the state’s Fiscal Year 2024 budget, according to language in the bill.
The bill, first introduced in May 2022, was championed by supporters who noted that 42.4% of foreign-born New Jerseyans over five years old speak English less than “very well" and that more than 150 languages are spoken in the state.
Ricardo Kaulessar covers race, immigration, and culture for NorthJersey.com. For unlimited access to the most important news from your local community, please subscribe or activate your digital account today.
Email: kaulessar@northjersey.com
Twitter: @ricardokaul
Friday, January 26, 2024
How to use real-time translation on the Samsung Galaxy S24 - TrustedReviews - Translation
Samsung recently announced the new Galaxy S series phones for 2024, including the Galaxy S24, S24 Plus and S24 Ultra.
The Galaxy S24 is packed with AI-powered features, including Live Translate for phone calls. The feature translates calls in real-time directly in the native phone app, allowing you to chat with someone in another language without downloading any additional apps or services.
Live Translate supports 13 languages at launch, including Chinese Simplified, English, French, German, Hindi, Italian, Japanese, Korean, Polish, Portuguese, Spanish, Thai and Vietnamese.
Keep reading to learn how to set up Live Translate and how to access the feature during calls.
What you’ll need:
- The Samsung Galaxy S24, S24 Plus or S24 Ultra
The Short Version
- Open your phone’s Settings
- Tap Advanced Features
- Tap Advanced Intelligence
- Select Phone
- Toggle Live Translate to On and select your languages
- Make or answer a phone call
- Tap Call Assist
- Tap Live Translate
How to use real-time translation on the Samsung Galaxy S24
-
Step
1Open your phone’s Settings
The first step is to set up Live Translate. Once the feature is set up, you can access it with a tap during calls by skipping to step 6 in this guide.
-
Step
2Tap Advanced Features
You might need to scroll to find this heading.
-
Step
3Tap Advanced Intelligence
This is where you’ll find settings for many of the new AI features on the Galaxy S24.
-
Step
4Select Phone
Or tap on the green phone icon.
-
Step
5Toggle Live Translate to On
Make sure the languages below match your language and the language of the person you plan to call. You might need to tap the Download icon next to a language to download it.
-
Step
6Make or answer a phone call
You can do this in the native phone app.
-
Step
7Tap Call Assist
This should be in the middle of the screen.
-
Step
8Tap Live Translate
Again, this should be in the middle of the screen.
-
Step
9That’s it
Bixby Voice will automatically translate between the two languages you selected.
Troubleshooting
You can also choose to mute one of your voices during calls so that you only hear the translation. You can also set specific language and voice presents for different contacts.
Live Translate supports 13 languages at launch, including Chinese Simplified, English, French, German, Hindi, Italian, Japanese, Korean, Polish, Portuguese, Spanish, Thai and Vietnamese.
You might like…
Last word on INDIA hasn’t been said yet. But the dictionary is running out of words - ThePrint - Dictionary
Why is Mamata Banerjee set on playing the lose-lose game? What does the two-term West Bengal chief minister hope to gain by breaking up the INDIA coalition even before it has really begun? There are now barely 90 days to go before the make-or-break 2024 Lok Sabha election. INDIA does not seem to have made any progress on thrashing out a strategy to take on the BJP. Yet, hours after Prime Minister Narendra Modi’s unprecedented show of strength in Ayodhya, Banerjee dropped a bombshell.
The Ekla Chalo bomb. No tie-up with the Congress in Bengal, she has declared. What Trinamool Congress does next will be decided after the election, she pronounced.
But why has Banerjee taken this extreme step?
BJP’s bidding
The CM has learned her politics the hard way. With no political godfather and no moneybags, she has pulled herself up to her current status by the bootstraps. Very few politicians in the country today can rival her mastery of electoral politics. Congress leader Jairam Ramesh has rightly said that INDIA is unimaginable without her. But it is also unimaginable that Banerjee pulled the rug out from under the coalition’s feet in a fit of pique without clear-eyed calculations. Banerjee can’t tell whether the move will benefit TMC or not.
What’s her arithmetic?
For those looking beyond the immediate drama, there are no easy answers.
The easy option is to look at what her direct political opponents in West Bengal—the Left and Congress—have to say. Both parties argue that Banerjee’s politics isn’t driven by arithmetic but history—a history of corruption by TMC leaders that has forced her to capitulate to BJP’s ED-CBI raj. This corruption, they allege, is not a figment of the ED-CBI’s imagination but a reality that has seen TMC leaders go in and out of jail over the last 12 years. Today, at least two senior ministers—Jyotipriyo Mallik and Partha Chatterjee—are “in” and need a lot of luck to get “out” unscathed.
If Banerjee wants to keep her flock, including nephew and TMC general secretary Abhishek Banerjee, out of jail, she has to do the BJP’s bidding. And the BJP has now bid her to break INDIA.
Also read: The Republic is dead and no point blaming BJP-RSS. We need a new political language
TMC’s huge worry
The more complex answers to the arithmetic question may lie in more math. Minority math, for starters. Minority votes in West Bengal stand at about 30 per cent and it goes to TMC en bloc. That is general knowledge. However, in the 2021 assembly election, a new-born party, the Indian Secular Front (ISF), founded by a ‘pir’ or religious leader Abbas Siddique made a tiny inroad. While Congress and Left each got less than five per cent of the vote share and not a single seat, the ISF snatched a seat from the TMC in Bhangar. The constituency is located in the South 24 Parganas district where Muslims account for a significant 30 per cent of votes.
Flash in the pan? Not quite, it seems. In panchayat elections last year, ISF spread its wings, even if with small wins, in North 24 Parganas, Howrah, Hooghly, and East Midnapore. It has big plans for 2024. Its lone MLA Naushad Siddiqui has been quoted saying he will contest from the Diamond Harbour constituency this time, the seat held by Abhishek Banerjee since 2014. ISF has clearly thrown down the gauntlet.
Months before the panchayat polls, the Congress’ victory at Sagardighi in an assembly election, had made TMC sit up and take notice. Sagardighi is in Murshidabad district, once the fief of Congress state chief Adhir Ranjan Chowdhury. He has repeatedly won Lok Sabha elections from Berhampore in that district but in 2021, TMC swept 20 out of 22 seats there—making him look shaky. Sagardighi is also minority dominated. TMC has since wooed the Congress MLA into its fold, but Sagardighi has caused unease in the TMC regarding cracks in the minority vote bank. ISF’s panchayat results added to that.
For TMC, it is a huge worry. Since 2011, and earlier, the minority vote has been its almost exclusive preserve. If this vote bank cracks, there could be serious trouble for the party. The Muslim vote nationwide is believed to be consolidating behind the Congress again. If that happens in West Bengal and ISF gets Muslim votes, the math could turn ominous for TMC.
Mamata Banerjee’s long game
The math has turned tricky for the BJP as well. Its 77-seat haul in 2021 is considerably whittled down, thanks to ghar wapsi by TMC MLAs who had switched sides during the assembly election. In Lok Sabha, the BJP’s 2019 tally of 18 seats is down to 17, with the party losing the Asansol seat. Even minus these losses, the BJP’s fortunes have not been on the upswing in the state. Leader of Opposition in the state Suvendu Adhikari and party state president Sukanta Majumdar are making noises but are unable to dispel the view that BJP is a fractious, faction-ridden and organisationally weak party, living under the hope that its Delhi bosses will come to the rescue.
But can TMC sit back and take it easy about BJP’s dwindling fortunes? No. Union Home Minister Amit Shah seems unrelenting in his West Bengal crusade. He has set a tall target for the BJP’s state unit: 35 seats in 2024 Lok Sabha. And Shah has been hopping in and out of the state. Even as Rahul Gandhi traverses north Bengal with his Bharat Jodo Nyay Yatra on Thursday, Shah will come to Kolkata on Sunday to run his own campaign.
Incumbency is a real thing, too, especially when charges of corruption are proving hard to dust off.
And finally, there is 2026. While 2024 Lok Sabha may seem to be the most important election of all time, it is not going to be the last. Assembly election is due in West Bengal in two years and there is no question that Banerjee is already drawing up a strategy to win it. She cannot afford to give INDIA partners Congress or Left even an inch in 2024 and risk any of them—including BJP and ISF—grab a yard in 2026. She has to ensure that the TMC perpetuates.
This way, perhaps, Banerjee’s lose-lose game in 2024 ensures that 2026 is a win-win. Banerjee knows if she breaks ranks, there will be nothing left of INDIA. Aam Aadmi Party (AAP) has followed in her footsteps in Punjab. The possibility of Bihar Chief Minister Nitish Kumar and his Janata Dal (United) doing a somersault is just not going away. For Banerjee, it is in her best interest to consolidate TMC’s position in West Bengal as much as possible, showing no quarter to any political rival in the state. Delhi can wait.
The last word on INDIA may not have been said yet. But the dictionary is running out of synonyms.
The author is a senior journalist based in Kolkata. She tweets @Monideepa62. Views are personal.
(Edited by Ratan Priya)
Thursday, January 25, 2024
A red flag for AI translation - POLITICO - POLITICO - Translation
Presented by
“Hallucinations” aside, today’s sophisticated chatbots can sometimes seem like magic — passing standardized tests with flying colors, or conjuring up multilingual poetry in the blink of an eye.
Well… depending on what language you speak. A recent paper awaiting peer review from a group of researchers at Amazon and the University of California, Santa Barbara found that chatbots’ linguistic skills might be threatened by ghosts from a past era of AI, raising significant questions about their ability to communicate effectively in lesser-used languages on the web (think regional dialects from Africa or Oceania). Analyzing a database of billions of sentences, they found that a huge chunk of the digital text likely to be hoovered into LLMs from those languages wasn’t written by native speakers but instead was crudely machine-translated by older AIs.
That means today’s cutting-edge multilingual models are training on very low-quality data, leading to lower-quality output in some languages, more hallucination, and potentially amplifying the web’s already-existing shortcomings and biases.
That’s obviously bad in its own right, but it raises a larger question about the future of generative AI: Is it doomed, as some have predicted, by the “garbage in, garbage out” principle?
I spoke today with Ethan Mollick, an AI researcher and professor at the University of Pennsylvania’s Wharton School, and asked him what he thought about the findings given his work on how people actually interact with AI models in a professional or classroom setting. He was skeptical that messy, photocopy-of-a-photocopy results like those the Amazon and UC Santa Barbara researchers found could lead to the “model collapse” that some researchers fear, but said he could see a need for AI companies to tackle language issues head-on.
“There are worlds where this is a big problem, and data quality and data quantity both matter,” Mollick said. “The real question is whether there’s going to be a deliberate effort, like I think Google has done with Bard, to try and train these models for other languages.”
Usually, large language models are trained with extra weight given to heavily-edited, high-quality sources like Wikipedia or officially published books and news media. In the case of lesser-used languages there’s simply less native, high-quality content in that vein on the web. The researchers found that AI models then disproportionately train on machine-translated articles they describe as “low quality,” about “topics like being taken more seriously at work, being careful about your choices, six tips for new boat owners, deciding to be happy, etc.”.
All it takes to determine what the “garbage in” to an AI model might be, then, is a quick web search. The “garbage out” is, of course, apparent from one’s interaction with the model, but exactly how it got made is less clear — and researchers like Mollick say the very size and sophistication of current AI models means that remains opaque to researchers for the moment.
“Even with open-source models, we just fundamentally don’t know” how, or why, certain AI models operate better or worse in any given language, Mollick said. “There are dueling papers about how much the quality versus quantity of data matters and how you train better in foreign languages.”
So, for those keeping score: Old, low-quality machine-translated foreign-language content does predominate in more obscure languages, reducing AI models’ fluency with them. But we don’t know exactly how this happens within any given AI model, and we also still don’t know exactly the extent to which AI development is threatened by training on AI-generated content.
Mehak Dhaliwal, a former AWS intern and current PhD student at UC Santa Barbara, told Vice’s Motherboard that the team initiated the study because they saw the lack of quality firsthand.
“We actually got interested in this topic because several colleagues who work in MT [machine translation] and are native speakers of low-resource languages noted that much of the internet in their native language appeared to be MT generated,” he said.
So what can actually be done about it? Brian Thompson, a senior scientist at Amazon AWS AI who is one of the paper’s authors and its listed contact, told DFD via email that he couldn’t comment. But he pointed to the conclusion of his fellow researchers that model trainers could use tools to identify and eliminate machine-translated content before it gums the model’s works up.
Both researchers and the data analysts fine-tuning these models are able to flag and classify data at an almost psychedelically minute level, meaning it should be no problem to at least attempt a prophylactic against bad translated content. Still, with the most sophisticated AI models like GPT-4 rumored to have roughly 1.8 trillion parameters, those scientists could have their work cut out for them.
franco files
Germany and France are pushing the European Union’s AI Act negotiations to their limit.
POLITICO’s Gian Volpicelli, Océane Herrero, and Hans von der Burchard reported on the back-and-forth for Pro subscribers today, as the two bloc heavyweights are pushing for more business-friendly strictures in the law ahead of a vote on its final text scheduled for Feb. 2.
“There is no final text,” a representative for the cabinet of French Economy Minister Bruno Le Maire told POLITICO Wednesday. “The regulation is still being negotiated, and that will result in another round of three-way negotiations.”
A particular sticking point is copyright, where France says the AI Act’s planned requirement that companies disclose the copyrighted material used in training AI would be an impediment to AI startups. Still, one diplomat familiar with the negotiations told POLITICO the law’s text is unlikely to be changed or blocked by force, saying France, Germany, and Italy are “isolated” in their preference for less stringent regulation.
good chatbot, good
Take a deep breath: The RAND Corporation says we don’t have to worry just yet about chatbots unleashing biological weapons.
In a report published this morning, RAND researchers say that “biological weapon attack planning currently lies beyond the capability frontier of LLMs as assistive tools” and that “LLMs do not substantially increase the risks associated with biological weapon attack planning.”
They ran a traditional controlled experiment, where one group of security experts planned an imaginary biological attack with LLMs and one planned an attack without them. The chatbots were of negligible help to the researchers.
Still, they say there might be room for, uh, “improvement” on that front: “It remains uncertain whether these risks lie ‘just beyond’ the frontier and, thus, whether upcoming LLM iterations will push the capability frontier far enough to encompass tasks as complex as biological weapon attack planning, or whether the task of planning a biological weapon attack is so complex and multifaceted as to always remain outside the frontier of LLMs,” they write in the conclusion.
The RAND researchers recommend scaling up various “red-teaming” exercises meant to detect malevolent AI activity before its use in the wild, and bolstering the research community around negative AI capabilities.
Tweet of the Day
THE FUTURE IN 5 LINKS
Stay in touch with the whole team: Ben Schreckinger ([email protected]); Derek Robertson ([email protected]); Mohar Chatterjee ([email protected]); Steve Heuser ([email protected]); Nate Robson ([email protected]); Daniella Cheslow ([email protected]); and Christine Mui ([email protected]).
If you’ve had this newsletter forwarded to you, you can sign up and read our mission statement at the links provided.