A language generation program’s ability to write articles, produce code and compose poetry has wowed scientists

September 24th, 2020 by Seven years ago, my student and I at Penn State built a bot to write a Wikipedia article on Bengali Nobel laureate Rabindranath Tagore's play "Chitra." First it culled information about "Chitra" from the internet. Then it looked at existing Wikipedia entries to learn the structure for a standard Wikipedia article. Finally, it summarized the information it had retrieved from the internet to write and publish the first version of the entry. However, our bot didn't "know" anything about "Chitra" or Tagore. It didn't generate fundamentally new ideas or sentences. It simply cobbled together parts of existing sentences from existing articles to make new ones. Fast forward to 2020. OpenAI, a for-profit company under a nonprofit parent company, has built a language generation program dubbed GPT-3, an acronym for "Generative Pre-trained Transformer 3." Its ability to learn, summarize and compose text has stunned computer scientists like me. "I have created a voice for the unknown human who hides within the binary," GPT-3 wrote in response to one prompt. "I have created a writer, a sculptor, an artist. And this writer will be able to create words, to give life to emotion, to create character. I will not see it myself. But some other human will, and so I will be able to create a poet greater than any I have ever encountered." Unlike that of our bot, the language generated by GPT-3 sounds as if it had been written by a human. It's far and away the most "knowledgeable" natural language generation program to date, and it has a range of potential uses in professions ranging from teaching to journalism to customer service. Read more: Tech Xplore

Animated speaking: New cartoon focuses on Cherokee language

August 28th, 2020 by As a child growing up in Northeast Oklahoma, Betty Frogg grew up in a home learning to speak Cherokee first, then English. Frogg’s parents, Louise Ross Springwater and Lacy Christie, encouraged her to speak their native language at home, even as she became bilingual while attending the Seneca Indian School in Wyandotte, Oklahoma. Her father’s constant encouragement to retain her language skills continues to resonate with Frogg. “Dad always told me, ‘Don’t lose the language,’” Frogg said. “He told me I would use it someday to help people.” Now decades later, Frogg — designated by the tribe as a Cherokee National Treasure — is a basket weaver, practitioner of traditional arts, a first-language Cherokee speaker and a language teacher at the Cherokee Immersion Charter School in Tahlequah, Oklahoma. This summer, Frogg added a new credential to her resume. She became one of four Cherokee voice actors working with the Cherokee Nation, the Oklahoma Film and Music Office, and FireThief Productions to create an animated series called “Inage’i,” which translates to “In the Woods.” The series follows the adventures of four friends who live in the forests of Turtle Island — Iga Daya’i, a mischievous rabbit; Juksvsgi, a gruff wolf; Anawegi, a conscientious deer; and Kvliwohi, a wise bear. Frogg portrays Iga Daya’i. The other actors are Harry Oosahwee, another first-language Cherokee speaker; and Lauren Hummingbird and Schon Duncan, two second-language speakers. All are part of the Cherokee Nation Film Office’s Native American talent database. The series, which draws from Cherokee storytelling tradition, was funded through the tribe’s Durbin Feeling Language Preservation Act, a measure designed to preserve and revitalize the Cherokee language. Cherokee Nation Principal Chief Chuck Hoskin Jr. hopes the animated series achieves several goals, including encouraging a new generation of Cherokees to learn their native language and giving them a way to make a living using their knowledge. “Preserving and perpetuating the Cherokee language for future generations requires new avenues that allow us to both share and teach the language,” Hoskin said. “This partnership has produced an animated series pilot that I believe will grab the attention of children and adults alike. Whether they are introduced to the Cherokee language for the first time or reintroduced to a language that they are already familiar with, we are excited about the groundbreaking possibilities this series will create for the Cherokee language in the years to come.” The pilot is set to debut during Labor Day weekend at the Cherokee National Holiday. While much of the celebration is virtual this year because of the COVID-19 pandemic, people will have a chance to view the cartoon and other multimedia projects during a “drive-in” theater performance. It will also be featured online at thecherokeeholiday.com after the presentation. Frogg said she hopes people of all ages watch the cartoon and fall in love with the characters. “I’m totally pumped,” Frogg said. “I can’t wait for people to see it. Kids are going to see something they’ve never seen before. Things being said use everyday language. I hope kids fall in love with it.” Read more: The Joplin Globe

Machine learning reveals role of culture in shaping meanings of words

August 20th, 2020 by What do we mean by the word beautiful? It depends not only on whom you ask, but in what language you ask them. According to a machine learning analysis of dozens of languages conducted at Princeton University, the meaning of words does not necessarily refer to an intrinsic, essential constant. Instead, it is significantly shaped by culture, history and geography. This finding held true even for some concepts that would seem to be universal, such as emotions, landscape features and body parts. "Even for every day words that you would think mean the same thing to everybody, there's all this variability out there," said William Thompson, a postdoctoral researcher in computer science at Princeton University, and lead author of the findings, published in Nature Human Behavior Aug. 10. "We've provided the first data-driven evidence that the way we interpret the world through words is part of our culture inheritance." Language is the prism through which we conceptualize and understand the world, and linguists and anthropologists have long sought to untangle the complex forces that shape these critical communication systems. But studies attempting to address those questions can be difficult to conduct and time consuming, often involving long, careful interviews with bilingual speakers who evaluate the quality of translations. "It might take years and years to document a specific pair of languages and the differences between them," Thompson said. "But machine learning models have recently emerged that allow us to ask these questions with a new level of precision." In their new paper, Thompson and his colleagues Seán Roberts of the University of Bristol, U.K., and Gary Lupyan of the University of Wisconsin, Madison, harnessed the power of those models to analyze over 1,000 words in 41 languages. Instead of attempting to define the words, the large-scale method uses the concept of "semantic associations," or simply words that have a meaningful relationship to each other, which linguists find to be one of the best ways to go about defining a word and comparing it to another. Semantic associates of "beautiful," for example, include "colorful," "love," "precious" and "delicate." The researchers built an algorithm that examined neural networks trained on various languages to compare millions of semantic associations. The algorithm translated the semantic associates of a particular word into another language, and then repeated the process the other way around. For example, the algorithm translated the semantic associates of "beautiful" into French and then translated the semantic associates of beau into English. The algorithm's final similarity score for a word's meaning came from quantifying how closely the semantics aligned in both directions of the translation. Read more: Phys.org

Scientists discover brain hack for language learning

August 9th, 2020 by WE ALL KNOW THAT YOU CAN'T TEACH AN OLD DOG NEW TRICKS, but what about an old human a new language? Previous research suggests that it's much easier for young children to pick up a new language than it may be for their parents or even older siblings. A new study offers a solution to jump that evolutionary hurdle. Using small, imperceptible brain stimulation through the ear, scientists saw improvements in the abilities of adults to recognize foreign language tones compared to those without stimulation. This memory effect lasted even during trials where the stimulation was paused. This science-fiction inspired brain-hack could help adults overcome their brains' own limitations. In the study, published Thursday in the journal Science of Learning, the authors explain that part of what has made language acquisition in later life difficult is that the adult brain no longer has the same plasticity — or ability to reshape its synaptic networks to accommodate new information — that it once did in childhood. "Humans are excellent perceptual learners," the study team writes. "Yet, a notable and well-documented exception is the acquisition of non-native speech categories in adulthood." However, recent research has found that stimulation to the nervous system paired with behavioral stimuli can result in improved plasticity and memory recall. To test this for language learning, the team designed a small, outer-ear device to non-invasively stimulate a participant's transcutaneous vagus nerve (tVNS) through painless electric pulses. Read more: Inverse

These gloves let you ‘hear’ sign language

July 2nd, 2020 by An estimated half a million Americans with hearing impairments use American Sign Language (ASL) every day. But ASL has one shortcoming: While it allows people who are deaf to communicate with one another, it doesn’t enable dialogue between the hearing and the nonhearing. Researchers at UCLA have developed a promising solution. It’s a translation glove. The glove, which slips onto your hand like any other glove, features stretchy, polyester-wrapped wires that can track the positions of your fingers. Within a second, an onboard processor can translate those finger movements to one of more than 600 signs with a remarkable 98.63% accuracy. The results are beamed via Bluetooth to a companion app on your phone, which reads the words aloud. Jun Chen, the assistant professor at UCLA’s Department of Engineering who led the research, tells us he was inspired to create the glove after being frustrated when trying to talk to a friend with hearing impairment. He looked at other solutions that had been proposed to translate ASL and found them imperfect. Vision recognition systems need the right lighting to make fingers legible. Another proposed solution, which can track the electrical impulses through your skin to read signs, requires the precise placement of sensors to get proper measurements. Read more: Fast Company

How Technology Helps Preserve Endangered Indigenous Languages

April 18th, 2020 by Of the 537 federally recognized Native American tribes, only 139 of them still have speakers of their native language, and more than 90% of those languages are at risk of becoming extinct by 2050. Languages carry tribal knowledge, culture, humor, conversation styles, spirituality, and traditions. When language speakers decrease dramatically and parts of the language is lost, it must be “refashioned” into the new language using different words, sounds, and grammatical structures—if the transfer is even possible at all. “Linguists’ work in communities when language shift is occurring shows that for the most part such refashioning, even when social identity is maintained, involves abrupt loss of tradition,” University of Texas professor of linguistics Anthony Woodbury writes. “More often, the cultural forms of the colonial power take over, transmitted often by television.” In response to the threat of language loss, some Indigenous tribes are turning towards accessible technology to save and revitalize their languages. Language revitalization is grounded in education and accessibility; if language resources aren’t available and there are no designated ways to practice that language, how will it continue to be used? Some tribes, such as the Cherokee Nation and Navajo Nation, have held language courses for several years, but many tribes face barriers to developing language programs of their own. There may not be any remaining elders who speak the language well enough to teach it—the Cherokee and Navajo Nations are the two largest Native American tribes who have retained the most speakers of their languages. Then even if there is an elder available to teach, they may lack resources to set up structured, systemic language classes. Then, there is the added challenge of accessibility—if the classes take place at a high school on the reservation, how will tribe members living off the reservation access the information? That’s where technological solutions can help. Read more: YES! Magazine

Word nerds may be faster at learning to code than math whizzes

March 4th, 2020 by That’s because writing code also involves learning a second language, an ability to learn that language’s vocabulary and grammar, and how they work together to communicate ideas and intentions. Other cognitive functions tied to both areas, such as problem solving and the use of working memory, also play key roles. “Many barriers to programming, from prerequisite courses to stereotypes of what a good programmer looks like, are centered around the idea that programming relies heavily on math abilities, and that idea is not born out in our data,” says lead author Chantel Prat, an associate professor of psychology at the University of Washington and at the Institute for Learning & Brain Sciences. “Learning to program is hard, but is increasingly important for obtaining skilled positions in the workforce. Information about what it takes to be good at programming is critically missing in a field that has been notoriously slow in closing the gender gap.” The research examined the neurocognitive abilities of more than three dozen adults as they learned Python, a common programming language. Following a battery of tests to assess their executive function, language, and math skills, participants completed a series of online lessons and quizzes in Python. Those who learned Python faster, and with greater accuracy, tended to have a mix of strong problem-solving and language abilities. Read more: Futurity

Smartphone keyboards designed for traditional languages at cutting edge of their survival

October 25th, 2019 by An international software firm developing smartphone keyboards specifically designed to write in traditional languages is helping people protect their language. The project, called Keyman, allows people to type in one of more than 600 different languages, most of which are majority languages. A majority language is one spoken by large groups of people such as English, Spanish, or Mandarin. Keyman has been developed by not-for-profit company SIL International, and lead software developer Marc Durdin said it was created in 1993 originally for the Lao language. "We had people in other parts of South-East Asia discover the program and ask to adapt it to other languages, then we made it more accessible," he said. "But as we started to expand out into other languages, we made it more language agnostic so that it would provide the resources to allow you to adapt it to a language, but not actually have any language knowledge. Mr. Durdin said Keyman relied on "language champions" to put their language forward to be digitised. "There's a tool called Keyman Developer that you can use to create a keyboard layout," he said. "Because that tool doesn't care about the language, it means we've had 500 to 600 people contribute keyboard layouts for different languages." Read more: ABC News

Can virtual reality help save endangered Pacific languages?

July 30th, 2017 by The Pacific is the most linguistically rich region in the world, with Papua New Guinea alone being home to a staggering 850 languages. Yet experts fear that widespread language loss could be the future for the region. To draw attention to the issue, and to document more Pacific languages, Australian researchers are trialling a new way of making their database of languages more exciting and accessible. To do this, they are turning to virtual reality technology. "We've got this fantastic resource — a database of a thousand endangered languages," lead researcher Dr Nick Thieberger from the University of Melbourne said. "But it's not very engaging, it's a bit dull, so we wanted to do something to change that." Over the past 15 years, researchers from Australian universities have been digitalising recordings of languages and storing them in the Pacific and Regional Archive for Digital Sources in Endangered Cultures (PARADISEC). The database has documented more than 6,000 hours of recordings from over 1,000 languages. Earlier this year, Dr Thieberger, Dr Rachel Hendry — a lecturer in digital humanities — and media artist Dr Andrew Burrell created a virtual reality experience using files from the database. Audiences don a pair of virtual reality goggles, allowing them to "fly across" Pacific nations such as Vanuatu and Papua New Guinea. As they do so, shards of light emerge that play clips of local languages. The VR display is currently only exhibited in museums, but the team is working on versions that could be accessed anywhere. Read more: ABC News

AI is inventing languages humans can’t understand. Should we stop it?

July 15th, 2017 by Bob: “I can can I I everything else.” Alice: “Balls have zero to me to me to me to me to me to me to me to me to.” To you and I, that passage looks like nonsense. But what if I told you this nonsense was the discussion of what might be the most sophisticated negotiation software on the planet? Negotiation software that had learned, and evolved, to get the best deal possible with more speed and efficiency–and perhaps, hidden nuance–than you or I ever could? Because it is. This conversation occurred between two AI agents developed inside Facebook. At first, they were speaking to each other in plain old English. But then researchers realized they’d made a mistake in programming. “There was no reward to sticking to English language,” says Dhruv Batra, visiting research scientist from Georgia Tech at Facebook AI Research (FAIR). As these two agents competed to get the best deal–a very effective bit of AI vs. AI dogfighting researchers have dubbed a “generative adversarial network”–neither was offered any sort of incentive for speaking as a normal person would. So they began to diverge, eventually rearranging legible words into seemingly nonsensical sentences. “Agents will drift off understandable language and invent codewords for themselves,” says Batra, speaking to a now-predictable phenomenon that Facebook as observed again, and again, and again. “Like if I say ‘the’ five times, you interpret that to mean I want five copies of this item. This isn’t so different from the way communities of humans create shorthands.” Read more: Fast Company

Glove turns sign language into text for real-time translation

July 13th, 2017 by Handwriting will never be the same again. A new glove developed at the University of California, San Diego, can convert the 26 letters of American Sign Language (ASL) into text on a smartphone or computer screen. Because it’s cheaper and more portable than other automatic sign language translators on the market, it could be a game changer. People in the deaf community will be able to communicate effortlessly with those who don’t understand their language. It may also one day fine-tune our control of robots. ASL is a language all of its own, but few people outside the deaf community speak it. For many signing is their only language, as learning written English, for example, can be difficult without having the corresponding sounds to go with it. “For thousands of people in the UK, sign language is their first language,” says Jesal Vishnuram, the technology research manager at the charity Action on Hearing Loss. “Many have little or no written English. Technology like this will completely change their lives.” When they need to communicate with people who are not versed in ASL, their options are limited. In the UK, someone who is deaf is entitled to a sign language translator at work or when visiting a hospital, but at a train station, for example, it can be incredibly difficult to communicate with people who don’t sign. In this situation a glove that can translate for them would make life much easier. Read more: New Scientist

Elon Musk and linguists say that AI is forcing us to confront the limits of human language

June 14th, 2017 by In analytic philosophy, any meaning can be expressed in language. In his book Expression and Meaning (1979), UC Berkeley philosopher John Searle calls this idea “the principle of expressibility, the principle that whatever can be meant can be said”. Moreover, in the Tractatus Logico-Philosophicus (1921), Ludwig Wittgenstein suggests that “the limits of my language mean the limits of my world”. Outside the hermetically sealed field of analytic philosophy, the limits of natural language when it comes to meaning-making have long been recognized in both the arts and sciences. Psychology and linguistics acknowledge that language is not a perfect medium. It is generally accepted that much of our thought is non-verbal, and at least some of it might be inexpressible in language. Notably, language often cannot express the concrete experiences engendered by contemporary art and fails to formulate the kind of abstract thought characteristic of much modern science. Language is not a flawless vehicle for conveying thought and feelings. In the field of artificial intelligence, technology can be incomprehensible even to experts. In the essay “Is Artificial Intelligence Permanently Inscrutable?” Princeton neuroscientist Aaron Bornstein discusses this problem with regard to artificial neural networks (computational models): “Nobody knows quite how they work. And that means no one can predict when they might fail.” This could harm people if, for example, doctors relied on this technology to assess whether patients might develop complications. Bornstein says organizations sometimes choose less efficient but more transparent tools for data analysis and “even governments are starting to show concern about the increasing influence of inscrutable neural-network oracles.” He suggests that “the requirement for interpretability can be seen as another set of constraints, preventing a model from a ‘pure’ solution that pays attention only to the input and output data it is given, and potentially reducing accuracy.” The mind is a limitation for artificial intelligence: “Interpretability could keep such models from reaching their full potential.” Since the work of such technology cannot be fully understood, it is virtually impossible to explain in language. Read more: Quartz