Keep it Clean. Please avoid obscene, vulgar, lewd, racist or sexually-oriented language.
PLEASE TURN OFF YOUR CAPS LOCK.
Don't Threaten. Threats of harming another person will not be tolerated.
Be Truthful. Don't knowingly lie about anyone or anything.
Be Nice. No racism, sexism or any sort of -ism that is degrading to another person.
Be Proactive. Use the 'Report' link on each comment to let us know of abusive posts.
Share with Us. We'd love to hear eyewitness accounts, the history behind an article.
Wednesday, May 29, 2024
India: Classic Tamil dictionary supported by Germany | National | insidenova.com - InsideNoVa - Dictionary
What does “friend dictionary” mean? Viral TikTok trend explained - Dexerto - Dictionary
The new TikTok “friend dictionary” trend is blowing up online, but what does it even mean? Continue reading for our full breakdown of the trend.
Almost every week, a new trend is formed on TikTok, these new types of videos, whether that be dances, sing-alongs, or jokes, all blowing up on the social media platform.
And while social media as a whole does at times have its own vocabulary, TikTok has taken this one step further with its new “friend dictionary” trend.
Continue reading as we explain all about this new viral content on the platform.
What is the TikTok “friend dictionary” trend?
TikTok has a strong following of users who fall into the Gen Z age range, something that the greater TikTok community has started to highlight with the new “friend dictionary” trend.
In essence, these videos are simple in design, with users asking people outside of the Gen Z age range if they know what certain expressions and phrases mean.
More often than not, the person being quizzed has no idea what any of the words mean, however, the quizzer then asks their friend or siblings who then immediately knows what the expression means. In other instances, these videos highlight the random expressions and “inside jokes” friends use with each other.
The trend has 40,000 posts under the friend dictionary tag, with some of the most popular videos amassing millions of views.
Content creator Chelsea Lefkowitz recently broke down what “friend dictionary” means in an article written by NBC, Lefkowitz stated, “I feel like it’s kind of a universal experience that most sisters or siblings or best friend duos can relate to.”
The TikToker and Instagram influencer added, “It’s kind of instinctual, especially for me and my sister. We have a shared background. It’s easy for us to develop inside jokes and references and we kind of have that unique way of communicating.”
And there you have it folks! Everything you need to know about the new “friend dictionary” trend taking over TikTok.
Tuesday, May 28, 2024
This AI Paper by ByteDance Research Introduces G-DIG: A Gradient-Based Leap Forward in Machine Translation Data Selection - MarkTechPost - Translation
Machine Translation (MT) is a significant field within Natural Language Processing (NLP) that focuses on automatically translating text from one language to another. This technology leverages large language models (LLMs) to understand and generate human languages, facilitating communication across linguistic boundaries. MT aims to bridge global communication gaps by continuously improving translation accuracy supporting multilingual information exchange and accessibility.
The primary challenge in machine translation lies in selecting high-quality and diverse training data for instruction fine-tuning. Quality and diversity in the data ensure that language models can generalize well across different contexts and languages. Without these elements, models may produce translations that lack accuracy or fail to capture nuanced meanings, limiting their effectiveness in real-world applications.
Existing research includes methods like in-context translation exemplar selection, prompt optimization, and decoding strategies to enhance machine translation performance. Notable models and frameworks include GPT-4, Bayling-13B, BigTranslate-13B, TIM, and NLLB-54B, focusing on instruction tuning and translation performance. These approaches leverage techniques to optimize translation accuracy and generalization, often relying on extensive datasets and sophisticated evaluation metrics such as BLEU, BLEURT, and COMET to measure effectiveness and improvements in language model translations.
Researchers from ByteDance Research have introduced a novel method named G-DIG, which uses gradient-based techniques to select high-quality and diverse instruction data for machine translation. The innovation leverages influence functions to analyze how individual training examples impact model performance. This method aims to improve data selection without relying on external models, thereby enhancing the quality and diversity of the training datasets.
The G-DIG method involves two main components: high-quality data selection and diversity enhancement. Researchers manually create a small set of seed data for high-quality data and use influence functions to identify training examples that positively impact the model’s performance. Specifically, they measure the response quality of each training sample with the influence score on test instances. To enhance diversity, they apply clustering algorithms to the gradients of training examples, ensuring various influences on the model. The gradient similarity is assessed using the Euclidean distance measure, and the K-means clustering algorithm is employed to group training data into diverse patterns. This two-step process ensures the selected data is high-quality and diverse, improving the model’s overall translation capabilities.
Extensive experiments on various translation tasks, including WMT22 and FLORES, demonstrated that G-DIG significantly outperforms existing data selection methods and achieves competitive results against state-of-the-art models. G-DIG performed better in both Zh → En and De → En translation tasks. For instance, in Zh → En translation, the G-DIG model consistently surpassed the random model across all metrics and dataset sizes. The COMET score for Zh → En translation improved by 1.7 with 1000 training examples and by 2.11 in BLEU on the FLORES dataset. In De → En translation, G-DIG improved BLEU scores by 2.11 and 1.24 on WMT and FLORES compared to models trained with randomly selected data. The researchers highlighted that models trained with G-DIG-selected data exhibited better translation quality and alignment with human expectations.
The research team successfully addressed the challenges of data quality and diversity in machine translation by introducing the G-DIG method. This approach leverages gradient-based data selection, enhancing the model’s performance without needing external quality assessment models. The study demonstrates the potential of G-DIG to improve translation accuracy and efficiency, paving the way for more advanced and reliable machine translation systems. Furthermore, G-DIG’s ability to select training data directly impacting model performance ensures that LLMs are better aligned with human instructions, making them more effective in real-world applications.
To summarize, ByteDance Research has introduced a groundbreaking method that addresses critical issues in machine translation, demonstrating significant improvements in translation quality through innovative data selection techniques. The G-DIG method represents a substantial advancement in the field, offering a new pathway for enhancing the capabilities of LLMs in various language translation tasks. This method’s success emphasizes the importance of high-quality and diverse data in training robust and accurate language models, ensuring they can meet global communication and information exchange demands.
Check out the Paper. All credit for this research goes to the researchers of this project. Also, don’t forget to follow us on Twitter. Join our Telegram Channel, Discord Channel, and LinkedIn Group.
If you like our work, you will love our newsletter..
Don’t Forget to join our 42k+ ML SubReddit
Nikhil is an intern consultant at Marktechpost. He is pursuing an integrated dual degree in Materials at the Indian Institute of Technology, Kharagpur. Nikhil is an AI/ML enthusiast who is always researching applications in fields like biomaterials and biomedical science. With a strong background in Material Science, he is exploring new advancements and creating opportunities to contribute.
Noah Webster and the Power of Words - Breakpoint - BreakPoint.org - Dictionary
On May 28, 1843, Noah Webster died in New Haven, Connecticut. Webster was a teacher and lawyer, an ardent supporter of the American Revolution, and a deeply committed Christian. In 1783, hoping to separate America from its cultural dependence on Europe, he published an American spelling guide that introduced most of the differences in spelling between American English and British English, many of which still exist today. He published a dictionary in 1806, which he expanded into the first full American dictionary in 1828.
Webster’s concern to create a uniquely American identity through the proper spelling and defining of words may seem odd in our image-driven culture, but that’s because we underestimate the importance and power of words. Words both reflect and shape culture in profound ways. For example, in his 1828 dictionary, Webster defined discernment as:
… also, the power or faculty of the mind, by which it distinguishes one thing from another, as truth from falsehood, virtue from vice; acuteness of judgment; power of perceiving differences of things or ideas, and their relations and tendencies.
And then, he added, “The errors of youth often proceed from the want of discernment.”
In contrast, today’s online Meriam-Webster dictionary defines discernment as “the ability to comprehend or grasp what is obscure” or as “an act of perceiving or discerning something.” Gone entirely is any sense of distinguishing between things, especially the role that moral sense plays in the process. The not-so-subtle shift reflects a cultural shift toward moral relativism.
Sometimes new meanings are merely reflected in our dictionaries. Other times, they are imposed. When they are, we get George Orwell’s Newspeak. For example, during the confirmation hearings for Supreme Court Justice Amy Coney Barrett, Democrats claimed her use of the term sexual orientation was offensive. It had not been considered offensive to that point but, in a blatantly political move the next day, Merriam-Webster changed the definition of the term in the dictionary to indicate that it now was offensive.
Sometimes, words change the world via new definitions, such as the forced use of masculine and feminine pronouns that are contrary to biological identity. Other times, the world is changed by new words, such as the invention of new pronouns for “non-binary” people. These are, in fact, attempts to use language as a tool to control thought. Our words can impose ideas.
Before complying with the linguistic demands of others, we’d do well to remember these words of Confucius, “When words lose their meaning, people lose their freedom.” For Noah Webster, freedom required simplifying the spelling of words so that Americans could be rid of the artificial aristocratic conventions of British English.
Today, cultural battles are most often over the definitions of words. Faithfulness to what is true will require holding on to old, criticized definitions, even in the face of attempts to impose new, artificial meanings on words. Specifically, the battle lies over the meaning of words like love, truth, marriage, family, freedom, equality, dignity, human, even man and woman, male and female, mother and father. As Confucius said, our freedom rests in whether these words lose their meaning.
Words are important, even worth fighting for. As the ever-quotable G.K. Chesterton said:
What is the good of words if they aren’t important enough to quarrel over? Why do we choose one word more than another if there isn’t any difference between them? If you called a woman a chimpanzee instead of an angel, wouldn’t there be a quarrel about a word? If you’re not going to argue about words, what are you going to argue about? Are you going to convey your meaning to me by moving your ears? The Church and the heresies always used to fight about words, because they are the only thing worth fighting about.
This Breakpoint was co-authored by Dr. Glenn Sunshine. For more resources to live like a Christian in this cultural moment, go to breakpoint.org.
'Friend dictionaries' on TikTok show how loved ones create their own languages - NBC News - Dictionary
For one pair of sisters, a brisk day isn’t chilly — it’s “Chilean sea bass.” And for one couple, they don’t receive deliveries — they “got a parcel.”
These phrases may be unusual to the average person, but to some duos they are common sayings.
Best friends, siblings and couples alike are sharing their inside jokes and shared lingo on TikTok. Dubbed “friend dictionaries,” and sometimes couple or sister dictionaries, these words reveal the secret languages people have with those closest to them.
The trend features videos of duos testing each other on their slang. Each person has to guess the term the other writes down for them.
TikTok creator Chelsea Lefkowitz posted five videos with her sister Amanda Paige guessing the unique slang they share, racking up over 11 million views total.
Viewers commented which terms were their favorites, including “nosetta” for a nosy person or “projectile committed” for plans they are unable to cancel.
“I feel like it’s kind of a universal experience that most sisters or siblings or best friend duos can relate to,” Lefkowitz said. “It’s kind of instinctual, especially for me and my sister. We have a shared background. It’s easy for us to develop inside jokes and references and we kind of have that unique way of communicating.”
Nicole Holliday, assistant professor of linguistics at Pomona College, said the trend exemplifies how “any community of people that have regular interactions with each other will come up with shorthand or references to previous events.”
“It’s establishing your closeness or your role in the community,” Holliday said. “I’m in on the joke. You’re in on the joke.”
Inside jokes and slang can also evoke feelings of nostalgia.
“It’s reminding you of the previous experiences that you’ve had together, which also sort of then facilitates the closeness of the relationship,” Holliday said.
It’s establishing your closeness or your role in the community.
Nicole Holliday, assistant professor of linguistics at Pomona College
People like the trend because it reminds them of how connected they are to their loved ones, according to Holliday.
Shared jokes and language require people to “know something about each other,” she said.
Reminders of connection and closeness are also likely appealing to audiences because of the ongoing loneliness epidemic and feelings of isolation following Covid-19 lockdowns.
“It makes them feel a sense of belonging with their family or friends,” Holliday said.
Jedson Tavernier and Jade Smith, who make couples content on TikTok, made two videos participating in the trend that received a total of over 19 million views. Viewers noted that they use funny voices and accents in addition to slang, which many related to doing with their own partners.
Tavernier said the videos are relatable and humanizing, which is refreshing on social media, where audiences can often forget that creators are people too.
“These types of videos that kind of break that wall and break that barrier down, where people are like, ‘Wow, they’re literally just like us,’ are really good for everybody,” he said.
Smith said she thinks people like the trend because it shows how being silly with your loved ones is a universal experience.
“Everyone has their own shared language and I think that makes you see everyone [as] really human,” she said. “I love that about it. I just think it’s so funny to see that everyone does what we do.”
Sunday, May 26, 2024
I ditched Google Translate for ChatGPT during my travels, here's why - Android Authority - Translation
Edgar Cervantes / Android Authority
If you’re a frequent traveler, Google Translate likely takes the top spot as the single most useful app on your smartphone. Communicating abroad can be difficult, after all, especially if it’s in an alphabet you’re not familiar with. But even though Google Translate has served me well for many years, it tends to fail at the most inopportune times. The solution? Using ChatGPT instead. It may sound clunky or unnecessary, but I assure you that the difference in translation quality is worth the jump from Google Translate to an AI chatbot like ChatGPT.
ChatGPT vs Google Translate: Why I’m switching for good
I could bore you with an academic test comparing Google Translate and ChatGPT in two or more languages I’m familiar with. But instead, I’ll show you a couple of the many real-world instances during my travels where ChatGPT has blown Google Translate out of the water.
First, some background. Google Translate has been around for longer than modern large language models. It relies on a system of vocabulary matching and statistical models. While it has picked up some machine learning tricks in recent years, ChatGPT and Google’s own Gemini model have far more language training under their belt. Not to mention, the latter have been specifically fine-tuned for conversation and are capable of understanding cultural nuances. All of these differences add up to a much superior experience.
Google Translate doesn't work as reliably across all languages.
The difference between Google Translate and ChatGPT became painfully obvious on my last trip to Vietnam. Vietnamese, unlike many of the world’s most spoken languages, is a tonal language with a complex system of diacritics. Miss one diacritic and the meaning of the word can change drastically. For example, “cÆ¡m cháy” translates to “burnt rice” whereas “cÆ¡m chay” is “vegetarian rice”. That’s just one missing diacritic, and it happens often when you’re trying to use an average smartphone camera to translate the world around you.
The result was that Google Translate would constantly mistranslate restaurant menus to the point of frustration. I eventually decided to switch to ChatGPT instead and was floored by the results. Take a look for yourself in this example where I sent a restaurant menu to both, Google Translate and ChatGPT:
Sure, Google Translate gets many menu entries right but it doesn’t understand nuance or context, so it has no idea whether it’s translating a novel or restaurant menu. This means it has trouble with several items and we end up with poor translations like “chemicals beans”. On the other hand, ChatGPT can identify a restaurant setting from the image or I can specify it in a follow-up message. This is invaluable as it can prevent obvious mistranslations like “dandruff” from showing up (yes, I’ve had that happen too). And as you can see in the screenshot above, ChatGPT had no problem working out the menu perfectly.
I’m not basing my preference for either tool on Vietnamese alone. I’ve had great success using ChatGPT in Japan too. Google Translate admittedly does a much better job with Japanese in general, but it would still stumble with packaged items and other labels. ChatGPT goes one step beyond serving as a simple translation tool in these cases.
Do I need to pay for ChatGPT’s superior translation?
Calvin Wankhede / Android Authority
It’s worth noting that you don’t need the paid version of ChatGPT to use its translation capabilities. The new GPT-4o model is now available to everyone, even if you don’t opt to pay for a monthly ChatGPT Plus subscription. Beyond the ability to upload images, the new model also offers higher-quality responses which naturally extends to translating foreign languages.
I’ve also had luck with ChatGPT’s voice conversation mode, which is also available for free in its basic form. However, GPT-4o’s faster and more emotive voice chat mode is locked behind the ChatGPT Plus subscription.
I’ve used the voice mode to translate foreign words vocally or ask for pronunciation advice. It’s an invaluable tool for language learning since ChatGPT can speak dozens of languages. To use it, simply download the ChatGPT app on Android or iOS and tap the headphones icon after logging in.
Of course, it doesn’t matter whether you use Google Translate or ChatGPT for simple one-liners. If you just need to ask for directions or translate between commonly spoken languages like English and French, chances are that Google Translate will serve you well enough. But for more obscure languages, especially those in a different alphabet, you’ll be best served by ChatGPT.
Google Translate vs ChatGPT: Where the old way still wins
Dhruv Bhutani / Android Authority
Unfortunately, even the best translation system in the world will fail if you don’t have an internet connection. As you may already know, ChatGPT doesn’t work offline, and you can’t download a language pack for translation as it’s possible within the Google Translate app. Modern AI chatbots require large amounts of computational power and our phones cannot generate responses on-device, at least not yet.
Luckily, eSIM phones and plans have become shockingly cheap over the past couple of years. It’s now easy to remain connected in most foreign countries for a few bucks, and you can often get a physical SIM card for even cheaper if you’re staying long-term. This gives you more than enough data to chat with ChatGPT and even send images for translation.
For the rare times when I lose connectivity altogether, I simply fall back to Google Translate. It’s rare enough in major tourist destinations and I’m not one to hike in the wilderness abroad, so it’s still worthwhile to use ChatGPT the majority of the time.
The only other downside to using ChatGPT instead of Google Translate is that AI chatbots aren’t as quick as plain old machine translation. This is especially true if you’re submitting images as multimodal AI is relatively new tech. Still, it’s a marvel any of the modern AI technology works as well as it does and I’m sure not complaining if the end result is this good.