Hacker News Reader

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 claims to have never said), or the nineties articles as to why no one would shop online.

3) Pimsleur is brilliant for developing basic conversational ability. It works really, really well.

4) The Pimsleur model could take you all the way to fluency, but it stops at a half-year of lessons for the most developed languages (and 2 weeks for some of the less developed ones). Developing four years of lessons at this level of quality is hard and expensive.

5) BUT: Moving on beyond Pimsleur, there is a range of tools which are passable. Things like Duolingo, Memrise, Supermemo, etc. can help build vocabulary. There are apps for watching movies for language learning (which adds subtitles, dictionaries, etc.) which can help build comprehension. Etc. Picking-and-choosing, you can get something which, while not as good as going there, is much better than classes.

6) This field is progressing super-quickly on the whole. What we have today is much better than five years ago, and we'll have much better things five years from now. I'd be surprised if apps weren't the best way to learn a language in not too long.

> Classes with other students are just about the least effective way to learn a foreign language.

Well, sure a private tutor is better, but that's a lot more expensive.

Without a teacher of some sort that can provide feedback, I have no idea how I could have learned elementary Mandarin (that is produce and listen to sounds radically different from my native English).

From what I've seen the ability for computers to give feedback on what you are doing wrong is far away from a trained human's.

> The Pimsleur model could take you all the way to fluency,

I've been looking over their website - it seems like it is just audio lessons and no tutor is provided? If that is correct, how would someone learn to correctly pronounce words in say Mandarin using this?

>> Classes with other students are just about the least effective way to learn a foreign language. > Well, sure a private tutor is better, but that's a lot more expensive.

In my experience, a private tutor really is not any more expensive than an classroom environment when you factor in the reduced number of hours needed. I have been taking language lessons online with 1-on-1 tutors (via Skype) through italki. You can easily find good native speakers with teaching qualifications for under $15 / hour. I find that I get more out of a dedicated hour with a tutor than half-day course in a classroom.

My brother and sister get individual Japanese lessons from a UCSC professor for 60 minutes a week. They apparently progress faster than the actual Japanese class. (I believe a normal UCSC class has 3-4 hours / week of classroom time.)

+1 for italki, being able to talk with a native speaker on demand for a low price is so awesome.

> 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 always ignore the feedback.

But I wouldn't say children don't get feedback at all. They get a lot of important feedback, such as whether the thing they wanted to happen when they spoke did happen, or whether the person they were talking to appeared to understand them. However, this kind of feedback is generally not the kind provided by a "teacher"; everyone automatically provides it.

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 understand the words.

In the Mandarin context, this means it is difficult to learn the ü sound - it just is perceived as something like "u" (resulting in inability to distinguish lü from lu). Tones are another category, where it is just plain hard for adults to distinguish words by tonality. Needless to say, production of the novel sounds is also incredibly hard.

Point being - it's difficult to bridge this gap - and I'd imagine near impossible for most without a teacher. I find Pimsleur's claim of learning a language "effortlessly" with a "near-native accent" - with only the aid of a CD - absurd in this context.

A source that goes into this: https://www.ncbi.nlm.nih.gov/pmc/articles/PMC2846316/

> Children however aren't handicapped by this problem -- they actually learn the phonemes of their language before they understand the words.

This is accurate, but it's worth pointing out that infants are born recognizing every potential phoneme, and what they learn to do is to ignore the differences between sounds that their language considers equivalent. They don't learn to make distinctions they weren't formerly able to make. (This supports the idea that learning foreign sounds is much harder for adults -- I mention it because most people intuitively believe that things happen the other way around.)

Adults do preserve the ability to learn to ignore the difference between two sounds, but that's much less useful than learning a distinction would be. :(

> Tones are another category, where it is just plain hard for adults to distinguish words by tonality.

English features tones pretty heavily, most prominently in the usually-obligatory tonal marking for yes/no questions, but interestingly also in the I-don't-know tonal sequence.

> I find Pimsleur's claim of learning a language "effortlessly" with a "near-native accent" - with only the aid of a CD - absurd in this context.

Agreed.

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.

>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 clarity, what I do with mispronunciations is break down the word in to phonemes. Yellow is a classic, that a lot of kids struggle with, current child subject has a propensity to say "lellow", though he gets it right more often than not now (age 3) -- I break that down as "yuh-ell-oh" and have him say other words that he can pronounce like "yes" and "yum". So I go "yes, yes, yes, yes, yellow" whilst he listens, then we do it together.

Most early language acquisition is done at home, and for us that has meant away from other parents so I'm not really sure how others do it. I have seen the "aw that's cute" reaction to mispronunciation in the wild, which I guess is a blocker for some.

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.

> I've been looking over their website - it seems like it is just audio lessons and no tutor is provided? If that is correct, how would someone learn to correctly pronounce words in say Mandarin using this?

Audio lessons vary widely in how they're designed. In the context of Pimsleur, it's an interactive back-and-forth.

Tape: "Tell her your flight is tomorrow."

[30 second pause. I say "你的航班是明天。"]

Tape: "你的航班是明天。"

[30 second pause. I correct "你的航班是明天," based on the feedback]

Tape: "你的航班是明天。"

[30 second pause. I repeat again]

And then the tape goes on.

That is feedback. It's prescriped, but it's incredibly effective. After I say something, I hear it pronounced back at me by a native speaker. That's 90% of the feedback a tutor or teacher would give you, only it's there for everything you say.

I found my pronunciation after a couple of months of Pimsleur to be dramatically better than people who've been in classes for a year or two. Indeed, when I tried stepping into a class, I left, since my pronunciation got worse.

That, combined with spaced repetition (which Pimsleur uses), combined with a half-dozen other research-based techniques, makes it ultra-effective.

> After I say something, I hear it pronounced back at me by a native speaker.

I imagine that being sufficient is going to vary a lot by individual learning styles and ability to mimic. I'm guessing you are rather good at mimicking sounds/music/etc.

> That's 90% of the feedback a tutor or teacher would give you, only it's there for everything you say.

I'm not so great at mimicking, so without someone telling me I'm actually pronouncing the words wrong, I'm just going to converge on an incorrect pronunciation. I cannot self-assess the accuracy; I need a tutor to point out that what I mistakenly believe is correct is not correct -- that is provide feedback to me. (I've tried similar exercises to what you describe and it just results in incorrect convergence)

This is especially an issue with Chinese where my native English brain naturally ignores tonality. Since the information is being stripped away, it's borderline impossible to self-ascertain whether my tones match or not.

Try it. You might be surprised. I thought I was really lousy at that sort of thing too.

It turns out the back-and-forth has a big impact.

That's not just me. Lots of my friends use it. Nice thing about CDs is you can get them from the library free too (or interlibrary loan).

I think the problem with Pimsleur is that the number of topics that you could become "fluent" in explodes once you get past a certain level of proficiency. I doubt that it could take you to fluency except in a very narrow range of topics due to the degree of repetition required.

> There are apps for watching movies for language learning (which adds subtitles, dictionaries, etc.)

I've been searching for such apps (for watching movies/audio-visual content). Got any names to share? Thanks.

I recently discovered https://languagelearningwithnetflix.com/

You get a nice overview of movies in a certain language available in your country. I don't use their Chrome extension though, so I can't say anything about that

I'm one of the two developers, thanks for the mention. :-)

We're working on YouTube support now for the extension.

Something else I've been working on: https://forum.language-learners.org/viewtopic.php?t=8699

It would be great to crowd-source data on whether Netflix's audio track matches the subtitles for a particular film. Even when the original audio language and subtitle language match, the subtitles often paraphrase the spoken dialogue, making it difficult to follow along.

In my investigations, this isn't too common, but I guess it depends on where in the Netflix catalogue (14,000 titles) you are digging. When Original language of film === audio language === subtitle language, things generally match up, in my experience. The subtitle accuracy for some less popular or older material can vary a bit though.

I'm the author of jijimaku, a free app to add dictionary definitions to subtitles if that helps.

>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.

And yet, hundreds of millions of people have learned a foreign language that way.

In this corner of Europe it's de rigueur for everyone to study 2 or more foreign languages during middle and high school, and (some very rich kids with private tutors aside) everybody does it at foreign language schools with classes of 10-15 other people. It still works just fine.

>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.

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.

This sounds amazing. I used Duolingo for about a year to learn Swedish. I got pretty good at reading it and hearing it, but never could speak it very well. Had I decided to learn French or Spanish I’d have a lot more opportunity to talk with other people, but I picked Swedish because it seemed interesting and easy for an English native to pick up. I would totally buy a VR rig just to do this.

Literally did the same last year with Swedish. Kind of amazing how easy it was to pick up vs French.

What kind of a time investment did it require on a day to day basis? In the last few weeks I've been eyeing Babbel and Duolingo as good options to get started learning a particular language.

Maybe an hour or so a day. If you're not talking with someone else the trick is to do it every single day. If I missed a day I lost a lot of momentum.

Also I'd recommend looking at Memrise. Their voice seems a lot better than duolingo's at least.

No need for VR.

busuu[1] have setup a network where users can send exercises to other members (native speakers of the language you are trying to speak) and they will correct and rate them. The network is very active, as soon as I submit my exercise, a couple of minutes later someone will have corrected it. (Premium plan though, I made a 3 months plan, hopefully it will be enough to get me on track).

I would suggest everyone to give busuu a try.

[1]: https://busuu.com [2]: https://imgur.com/a/3PrxDBH

Agreed on immersion being critical.

You also for languages distant from ones you know really need a teacher (which is a problem with Duolingo), at least to break into the elementary levels.

> You speak the words, but the vendor doesn't understand. Your tones are off.

The immersion is critical to having repetitive practice and exposure, but few adults (at least I'm not one) will get this far (that is being capable of extending knowledge via immersion) without the initial aid of a teacher. e.g. Learning to hear/speak sounds that don't exist in your native language requires both extensive instruction and feedback more detailed than just "I don't understand".

There is an AR translator app "Translate Now" that translates objects and text in Augmented Reality. I think it can create an immersive language learning experience for students to see translations placed in the "real world" around them. This is the appstore link: https://itunes.apple.com/app/apple-store/id1348028646?mt=8

I do believe that language learning can improve a lot using new technologies such as AR and VR. VR experiences as you described above can really impact how students aquire new skills.

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

There's no reason that needs to be in VR right?

Needs? No.

But immersing yourself in the environment in which you’ll use the language should eventually help with cued recall.

I remember an old game called Mean City (https://www.mobygames.com/game/mean-city-learn-english-or-di...) that tried to do something similar (minus the VR and the voice parts). There were probably others.

Find a MUD in your target language.

https://www.reddit.com/r/MUD/comments/bhxejm/muding_for_span...

I think Lingotopia was supposed to be something like this (though not VR).

https://www.lingoludo.com/lingotopia/

I really wish this VR existed. Would be great.

This would be close to AGI.

https://www.alelo.com/products-and-solutions/