Comment by wodenokoto | Hacker News Reader

Comment by wodenokoto | Hacker News Reader

"Lingvist was founded by people who helped discover the Higgs boson at CERN as well as the people who built the core technology at Skype"

Neither of which are skills that have anything to do with language. Not a single word about scientific testing of this audacious claim that you can learn a language in 200 hours.

At least Duolingo actually tested people against a regular college course and came out on top, so you have that to go on as a user.

They do specify what learn means - learn 3000 words. Is that recall? Recognition? Use in sentence? Actually understand in a sentence? In 200 hours?

I call bullshit.

If you look at the features page (how it works) you'll see that they claim to be the first electronic spaced repetition system ever in the entire world ever.

What about SuperMemo, Anki, Duolingo, Memrise, and hundreds of other SRS services and software?

Save your money, build an Anki deck or get in on Duolingo. Talk with a native speaker. Work your ass of. This is no silver bullet.

Duolingo sucks though. So does Memrise. And you can build an anki deck, sure — but not all decks are created equal. If you don't make a frequency list from a large corpus, your Anki deck sucks too.

Duolingo and Memrise seem to have the philosophy, "There are many words in the language you're learning. Here are some of them." This is really bad.

Because word frequencies follow a Zipfian distribution, if you learn the top 1,000 words in a language you're a long way forward. If you learn an arbitrary 1,000 sample of words in a 40,000 word vocabulary, you're much further back.

These guys seem to have this as their specific value proposation. I was going to do exactly this myself. I'm a computational linguist who's recently moved to Germany. I got about 2/3 of the way through the Duolingo course, but it's basically a waste of time. They taught me the words for "duck" and "ceiling" and "handyman" as part of the first 1,000 words in the course.

Interesting point. As I noted elsewhere, I do agree that there is room for improvements in this space. I admit that my post is mostly fuelled by being annoyed with their marketing copy.

I think you need to be more careful about what are useful words. I remember my Japanese text book introduced the word for alien fairly early, and I thought it was stupid. Once I came to Japan, I realised if I had learned that word, I could say more funny things in my very limited conversations.

When I was studying in China, we were introduced to the word for diarrhea and how to describe the texture of our stool within the first 2 months. Hardly top 1.000 words, but if you've ever lived in China, you realise it can come quite in handy!

Frequency lists are often heavily influenced by newspapers and that is probably not representative of the things you will need in daily life or conversations. I have wondered if subtitles are more representative of normal conversations, and there are some research into building corpus from that, but I'm not really sure how you test the utility of your corpus.

Newspapers are likely a bad choice, yes. I've thought about this a bit. Subtitles seem good, and there are corpora of casual speech.

I plan to heavily bias towards verbs. Nouns are much more long-tailed, and their frequency is "bursty" --- if the noun is relevant to the situation, it'll be used a lot. So you can get them when you need them. Otherwise you can always use a pronoun, or be vaguer.

Tangential, but...

> build an Anki deck

This is a giant pain in the ass and the public decks generally seem to be low quality (and painfully out of sync with whatever else I'm studying). I'd happily pay Duolingo for an Anki (or equivalent) deck they match to their lessons.

Can't you export your wordlists from duolingo to anki?

Also, decks such as core2000 + dictionary of basic grammar + remember the kanji or kanji damage, will give all the highest quality content I've seen to learn Japanese. You will still need some guidance from a teacher, especially to get started, but no matter what you do for language learning, you will need more than one avenue.

The spoon fed Chinese deck is way better than Rosetta Stone.

But yes, there is room for curated content in this space, I just got really annoyed with the tone of the marketing text and went of on a bit of a rant. At least they didn't claim to make you learn like a baby :)

Your point is valid.

I couldn't find a way to export the word list from duolingo. I may have overlooked something, but the best I found was to manually type up the vocabulary. (Copy-paste from the vocab page isn't very useful because it doesn't capture conjugations or articles.)

>Neither of which are skills that have anything to do with language.

Because nothing says "monoculture" and "single language" like scientists from all over the world converging outside Geneva to build the world's biggest atom smasher!

Fair point. But why focus on the physics aspect and not the communication aspect in the copy?

Reply on news.ycombinator.com