Hacker News Board Hacker News, browsed like a classic message board
Live data from Hacker News

500 days of Duolingo: What you can and can’t learn from a language app

nytimes.com

Post reply on HN
Print view View original

191–200 of 221 posts

Re: 500 days of Duolingo: What you can and can’t learn from a language app

#191

Post by hiharryhere »

bit_logic wrote: post #119

Learning a language requires immersion. That's why classes with other students are more effective than flashcards. VR has potential to really make a difference here. Imagine you load a VR level. It's a busy outdoor market in Taipei. In your inventory you have some cash. NPCs with AI and chat logic are around you interacting. Your goal is to buy some bread and get to the subway. You see a vendor with some bread, but d…

There's a guy building this exact VR thing in a startup program I'm a part of. Looks incredible: http://virtualnihongo.com/

Re: 500 days of Duolingo: What you can and can’t learn from a language app

#192

Post by barry-cotter »

pbhjpbhj wrote: post #176
thaumasiotes wrote: post #159

Earlier quoted context omitted.

> Without a teacher of some sort that can provide feedback, I have no idea how I could have learned elementary Mandarin One of the points made in The Language Instinct is that children pick up their languages without using any explicit feedback. That is, people in the child's life who can already speak are liable to ignore errors in the child's speech, and when they do point out errors, the child will essentially alw…

>That is, people in the child's life who can already speak are liable to ignore errors in the child's speech, and when they do point out errors, the child will essentially always ignore the feedback. // Interesting, I've some previous and am currently in a situation where I actively correct a child learner, and they respond well to that correction wrt language, less so in behavioural aspects unfortunately ;o) For cla…

As your experience illustrates this works well but it’s not necessary. Breaking down words into phonemes is not something people from illiterate cultures do normally. It’s not excessively difficult to teach by any means but it’s also not something that comes naturally as you’ve no doubt seen when teaching children how to sound out words when learning to read. Teaching consonant vowel consonant words (cat, rat, rag, bag, tag, tug, bug etc.) if you sound out a word phoneme by phoneme they’ll often get one wrong when they put it together, usually the vowel.

In the environment of evolutionary adaptedness language acquisition would not primarily occur in a nuclear family, children would spend a lot of time around not just their immediate family but also with other nursing mothers and children close in age and as they get older with children close on age more generally. They pick up pronunciation from the environment and all the other elements of a language too. This is why children will reject a language spoken only by one person in their environment starting around age three[1] and how you get distinct school/social class accents, like international school English or public school English in the U.K.

[1]http://www.todayifoundout.com/index.php/2012/08/a-man-once-t...

This guy spoke only Klingon to his son for the first three years of his life. The same rejection occurs with natural languages if only one person speaks it to a child so if you want your child to grow up bilingual they better hear your language from other people too.

Re: 500 days of Duolingo: What you can and can’t learn from a language app

#193

Post by yorwba »

usaar333 wrote: post #171
thaumasiotes wrote: post #159

Earlier quoted context omitted.

> Without a teacher of some sort that can provide feedback, I have no idea how I could have learned elementary Mandarin One of the points made in The Language Instinct is that children pick up their languages without using any explicit feedback. That is, people in the child's life who can already speak are liable to ignore errors in the child's speech, and when they do point out errors, the child will essentially alw…

Agreed. By "people" above, I should have clarified "adults". The issue is that it is very difficult for adults to learn new phonemes (individual units of sound that distinguish words from each other) - an adult hearing a novel phoneme tends to map it to one existing in their native language. Children however aren't handicapped by this problem -- they actually learn the phonemes of their language before they understan…

Wikipedia has https://en.wikipedia.org/wiki/Help:IPA/Mandarin and https://en.wikipedia.org/wiki/Standard_Chinese_phonology Similar pages exist for many other languages. If you study the descriptions of the individual phones carefully, you should be able to consciously position your tongue and mouth to produce the correct sound.

E.g. the ü sound is https://en.wikipedia.org/wiki/Close_front_rounded_vowel , which means that it only differs from https://en.wikipedia.org/wiki/Close_front_unrounded_vowel (the English ee sound) by the lips being rounded instead of unrounded. English also has rounded vowels, for example https://en.wikipedia.org/wiki/Close_back_rounded_vowel (the oo sound). So you can learn to consciously pronounce ü by pronouncing ee and then rounding your lips as when pronouncing oo while keeping the tongue in the ee position.

Most teachers don't know enough about phonology to give that kind of explanation. My Mandarin teacher certainly didn't. I can only credit Wikipedia with helping me perfect my pronunciation to the point where I can be mistaken for a native speaker on the phone.

Re: 500 days of Duolingo: What you can and can’t learn from a language app

#194

Post by SquareWheel »

netrikare wrote: post #188
Mirioron wrote: post #2

I envy the author's persistence. 500 days of working on a task without skipping a day is impressive.

A lot of people have impressive streaks - https://duome.eu/

I don't think there's any way to verify these are legitimate streaks though, and haven't been paid for. The whole concept really fell apart after Duolingo started monetizing that aspect of the app.

Re: 500 days of Duolingo: What you can and can’t learn from a language app

#195

Post by reitzensteinm »

wallflower wrote: post #81
reitzensteinm wrote: post #15

Frustration with Duolingo and Memrise when learning German led me to build a tool that could generate sentences in German and English simultaneously, while obeying the grammar rules. In my opinion, the handful of hand crafted sentences they provide bundles way too much complexity together at random. In a traditional text book, you'd get a dozen questions that are very similar, changing one or two aspects of a sentenc…

> So, I built a ClojureScript app that did just that. Sitting down and churning through hundreds of those allowed me to build an intuition for grammar rules that I knew on paper but would mess up when attempting to use them. If you did not use an existing rules engine, would you please consider open sourcing what you have done to help others who are considering building something similar? Or describe more in depth ho…

I had a bunch of helper functions, but my view was that there's no way to solve it in general; building exercises would always involve writing annotations via a DSL.

You wouldn't be able to avoid hand crafting each sentence structure (like you were writing a textbook), but the advantage comes from being able to parameterize the generation of tests. You define where the holes are and what can go in them for a particular exercise, and the combinatorial nature of it yields thousands of examples, with tunable complexity, allowing you to crunch on a topic with enough variation that you still have to think and not get bored.

Ich mag rote Bücher, aber er grüne besser ist denkt.

It's been two years so es ist viellicht nicht richtig!

I don't think if you solve it this way, my solution would really be a helpful starting point.

I didn't have a plan for moving past the first year or two of learning, I think you could learn a ton of basics this way, but once language gets more subtle and varied you'd probably need to do something else entirely.

Also I don't know if this would work with other languages. German is extremely regular.

Re: 500 days of Duolingo: What you can and can’t learn from a language app

#196

Post by rangerpolitic »

bit_logic wrote: post #119

Learning a language requires immersion. That's why classes with other students are more effective than flashcards. VR has potential to really make a difference here. Imagine you load a VR level. It's a busy outdoor market in Taipei. In your inventory you have some cash. NPCs with AI and chat logic are around you interacting. Your goal is to buy some bread and get to the subway. You see a vendor with some bread, but d…

VR+AI will not make a significant impact. I am a former language educator, and I've been through this before... and before... and before.

It follows a hype cycle. The idea is promoted. People get excited. It comes out. It fizzles. Later, educators rediscover the idea. They implement it. It has a small positive impact for small number of learners in a small number of cases.

To give some examples: - Native English Speaking Teachers - English immersion villages in South Korea - English robots in South Korea and Japan - The Audio-lingual Method - Berlitz Method - Rosetta Stone Immersion Based Learning Method - MUDs, MOOs, MUSHs - MOOCs - Blended learning - Badges - Gamification - Second Life

There's no reason to think VR+AI will be any different. It's simply what's hot right now.

Re: 500 days of Duolingo: What you can and can’t learn from a language app

#197

Post by Barrin92 »

gibba999 wrote: post #140
bit_logic wrote: post #119

Learning a language requires immersion. That's why classes with other students are more effective than flashcards. VR has potential to really make a difference here. Imagine you load a VR level. It's a busy outdoor market in Taipei. In your inventory you have some cash. NPCs with AI and chat logic are around you interacting. Your goal is to buy some bread and get to the subway. You see a vendor with some bread, but d…

I found the article to be BS. My experience: 1) Classes with other students are just about the least effective way to learn a foreign language. You just reinforce each others' bad accents, grammar mistakes, etc. 2) Duolingo is okay for some things, but not for others. But discounting all apps after having tried one or two, though, is a lot like the famous Gates quote: "640k should be enough for anybody" (which he cla…

>1) Classes with other students are just about the least effective way to learn a foreign language. You just reinforce each others' bad accents, grammar mistakes, etc.

In almost all cases the primary reason for learning another language is conversation. The richness of interaction trumps the fact of picking up an accent or grammar errors.

I learned French by living in France for half a year, and was relatively decent at it at the end of my stay. Over the last few years I've tried to learn several languages by book or through apps, and progress is significantly slower, maybe as much as ten times slower.

There's so much richness in real world interaction, including visuals, sounds, connecting words and phrases to specific tasks, it's impossible to replace with an app.

Re: 500 days of Duolingo: What you can and can’t learn from a language app

#198

Post by kilroy123 »

msvan wrote: post #7

The best way to learn a language is to need to learn it. If you come in contact with a foreign language daily, it becomes much easier to both practice and motivate yourself to continue to practice. As someone who has spent thousands of hours learning a foreign language, I've come to the conclusion that I will never do it again unless I have a proper reason for it. Might seem obvious, but it's worth pointing out.

As a "gringo" who lived in Mexico for a few years, I completely agree.

Re: 500 days of Duolingo: What you can and can’t learn from a language app

#199

Post by DrPhish »

Redoubts wrote: post #158
akg_67 wrote: post #90

Earlier quoted context omitted.

Duolingo is not good for learning East Asian languages. For Japanese, it is particularly bad. Try LingoDeer instead.

Why?

For one, the accents of the speakers they are using are humorously bad for Japanese specifically

Re: 500 days of Duolingo: What you can and can’t learn from a language app

#200

Post by barry-cotter »

reitzensteinm wrote: post #15

Frustration with Duolingo and Memrise when learning German led me to build a tool that could generate sentences in German and English simultaneously, while obeying the grammar rules. In my opinion, the handful of hand crafted sentences they provide bundles way too much complexity together at random. In a traditional text book, you'd get a dozen questions that are very similar, changing one or two aspects of a sentenc…

Once you’re past a basic level you’re at least as well off using Anki cloze deletion cards for this. You take a news article or short story and break it into chunks and the app will show it to you with bits missing and you have to guess/remember what’s in the missing parts.

So this is the source card with four possible deletions.

[Alle glücklichen Familien_c1] [sind einander ähnlich_c2]; [aber jede unglückliche Familie_c3] [ist auf ihre besondere Art unglücklich._c4]

And this is one question generated from that card.

[...] sind einander ähnlich; aber jede unglückliche Familie ist auf ihre besondere Art unglücklich.

You could do the same with a grammar textbook but if you use authentic texts I find it more enjoyable. If you want known good translations project Gutenberg has many different translations of classic works and aligning parallel corpora is a solved problem in natural language processing. If you want simple sentence structures you could start with children’s books.

Post reply on HN