The view of African languages from low-Earth orbit: www.linkedin.com/posts/clear-...

"Language is from people and it belongs to people." Computational Linguist and NLP expert, Alp Öktem, who was satisfied to see indigenous people as active participants in hashtag#LT4All2025 telling big tech to watch out for neocolonial ways of working. www.linkedin.com/posts/clear-...

Since the president has signed an EO making English the official language of the United States, here is the first non-English copy of the Declaration of Independence, printed in German just days after July 4, 1776. We have always been multilingual. cdm16274.contentdm.oclc.org/digital/coll...

This is TODAY - still 4 hours to register. You are invited to a Zoom meeting. When: Feb 28, 2025 02:00 PM Nairobi zoom.us/meeting/regi...

The language technology platform for Africa that ACALAN is questing to implement, being revealed at the UNESCO #LT4all conference (Language Technology for All) in Paris. #indigenouslanguages

Experts and stakeholders from around the world are meeting in Paris today regarding Language Technology for the billions who speak languages other than the usual suspects. Today's program at UNESCO #LT4all conference www.lt4all2025.eu/full-program... #indigenouslanguages

Recalling this rather glowing BBC piece from a few years back: www.bbc.com/news/technol.... Tech jobs in Africa are great when they have concern for fairness and well being, not when they are the cheapest way to extract max human labor. kamu.si/introducing-kamfupi - I hope we're better in Nairobi!!

Posted this on our #AI / #machinetranslation / #MT -focused account...

AI Spot Check: watch ChatGPT (o1) way OVERTHINK its response to the prompt: Changing one character per line, please transform SUSANCOLLINSOFMAINE to CONTEMPTIBLEENABLER kamu.si/susan-collin... #ai #chatgpt #artificialintelligence

Cheetah, an open-source toolkit developed by UBC-NLP, focusing on Natural Language Processing for African languages. Cheetah works *experimentally* in 517 African languages, a step on the road toward equitable NLP for non-lucrative languages. www.linkedin.com/posts/clear-...

🤔 What’s Interesting in Low-Resource Language Tech? Aimee Ansari discusses Nwulite Obodo Open Data License, a license for sharing African language datasets. Developed by Data Science Law Lab at the University of Pretoria, with others in South Africa and Kenya. www.linkedin.com/posts/aimee-...

AI Spotcheck: ➡ ChatGPT nailed it. ➡ DeepSeek failed it. DeepSeek went with MUSA, the Make Up Stuff Algorithm, which you'll want to read about at kamu.si/tyb-musa #machinetranslation #AI #MT #NLP #LLMs #deepseek #chatgpt #openai

First test of DeepSeek for a basic translation, using an authentic-text tweet with a common English expression, to another lucrative language. Major fail.

The 100th episode of Lingthusiasm is a fun race through the linguistics landscape. Advice learned the hard way - probably better not to listen on 1.7x speed, while in the final km's of a half marathon. Their hundred factoids do flash past rather quickly.

How about anachromojis? 💾 = save ⏩ = fast forward 📽️ = movie/ video 🗻 or ❄️ = snow - increasingly becoming a historical relic (kamu.si/whither-the-...) 🌃 = night stars replaced with Starlink satellites

Business Daily Africa: Harnessing power of language, tech to boost SME digital trade kamu.si/language-tec...

The African Languages group on Facebook: www.facebook.com/groups/46760...

Mostly for Africans... Are you positioned to go through the language list for your country and find errors in what is included? Languages might be • missing • double-listed with alternate names • listed with an alternate name instead of the primary • not a local mother tongue kamu.si/country-grid...

Nicely conceived and produced video about learning a language out of the mainstream kamu.si/welsh-learning

The disastrous fires in LA were, of course, predicted. Still, the world remains complacent to climate change because the "smart people" will figure it out. Except 🅰️ They have figured it out but 8 billion people are not ready to listen, and 🅱️ They haven't figured it out, e.g. they still drive and fly

New word for the disasters to be unleashed by the incoming Fourth Reich. Feel free to share liberally.

AI Spot Check: Caught ChatGPT CHEATING! YouTuber @albertatech posted a video on why LLMs can't count the Rs in strawberry Suddenly, it gives the right answer. BUT that's obviously a human intervention - my followup shows that the inherent process problem remains youtube.com/shorts/7pQrM...

Cleopatra lived closer in time to 'yeet' being in the dictionary than to the construction of the Great Pyramid of Giza.

Tanzania has a population of 62,000,000. 0.09 dental personnel per 10,000 people. That is, about 1 dentist for every 6 million people. My first thought on looking up the stat was, "It can't be THAT bad, can it?"

Regular Chinese (not scholars) - how tough is their language to write? kamu.si/chinese-char... Kamusi actually has an approach to a uniquely effective logical dictionary to locate Chinese terms without knowing the characters to write. But no funding prospects - can share the link if ur interested

Oh, this is fun! In 30 seconds, they'll analyze your English and guess where you're originally from. Give it a whirl - did they get you right? start.boldvoice.com/accent-guesser (We're not endorsing the product they're selling, though would be curious to hear from people who've used it.)

A first attempt to list every language spoken in Africa, per country. Please be brutal in finding things to fix! kamu.si/country-grid...

1st CEFR criterion to rank as A1 Basic language proficiency: Can understand and use familiar everyday expressions and very basic phrases aimed at the satisfaction of needs of a concrete type. How would you form a similar objective for an LLM that is training on a new language? kamu.si/cefr-levels

I'm looking for input on a new metric for LLM evaluation, based on how we rank human proficiency via CEFR. I'll post each human criterion & ask you: what is the computer task equivalent at that level? My basic premise is, metrics like BLEU are vapid ##s, we need something better kamu.si/cefr-levels

Even though we associate language death with modern times, languages have been dying out in Late Antiquity too. Sometimes (well, most of the time) we learn about how it happened in quite unusual ways. A little thread 🧵 1/ Isidore, Etymologies, bk. IX.1 On languages St. Gallen, Cod. Sang. 231

Persnickety question, what is a language. A bit of clarification at kamu.si/7000-languages

Technocolonialism at its finest. Here's a very real cautionary tale about what happens when AI is let loose on indigenous languages without serious human oversight: anishinabeknews.ca/2024/10/22/f...

Important perspective on the purposes and inadequacies of language technology for smaller languages. (You can watch on 1.75 speed, we won't tell anyone.) www.youtube.com/watch?v=YcU_...

The book Google doesn't want you to read

managed to customize our Bluesky handle, so we're @kamusi.org going forward

Fantastic illustration of the basic concept of the unfunded temporal dimension planned for the Kamusi data matrix, the "4" in Kam4D. kamu.si/kam4d-time - if you are not taken right to the relevant section (browsers work funny), look for "time" in the table of contents, or scroll

I like the term “loanword” because it implies the word will need to be returned at some point. “Café” is due back to the French in three days. The fines on “raccoon” have really been piling up with the Algonquin so make sure to return that this week

Pronouncing "banal" like "canal". Or either those without their first letter.

It says too much that Biden only manages to land symbolically in Africa at 11:58 pm in his presidency. Message: I care? This "gift" link to the NYT article will expire in 2 weeks: kamu.si/biden-africa...

A good way to frame the current state of AI for most marginalized languages. One can't pretend you'll get any sort of useful AGI without a heck of a lot of effort, but it's worth dipping a toe in the water. The CEFR scale is a helpful comparative metric!

kamu.si/%F0%90%AB%95... - with the unique software we developed at Kamusi Labs, did you know you can actually make URLs with Manichaean script (or hieroglyphs, or emojis, or any Unicode)? Check out kamfupi.link , but only if you're ready to be blown away.

Languages don't go extinct because they are not rich enough to convey the full range of the thoughts of their speakers. They do go extinct because their speakers are not rich enough to stand up to external pressures, whether outright oppression or the kids looking for jobs in a dominant tongue.

Monday. Rage and courage to all

Good opening post for our new kamusi.org Bluesky account? youtu.be/dqiypMz5xGo?...