10 ms·

An experiment in teaching a language without relying on translation

74 points by toto444 6y ago

zrkrlc 6y ago
I feel like this will only work if you can read hiragana and katakana already.
- maweki 6y ago
  Not being able to pronounce or hear the words is a problem. Associating the words just with the symbols seems very taxing to the visual parts of the brain. Also using the voice/vocal/speaking parts of the brain seems advantageous.
  - tsimionescu 6y ago
    Not to mention, even if you could manage to remember the symbols that way, it would still be next to useless, since you still couldn't SPEAK Japanese at all. The site creator has stated elsewhere though that the site is intended for people who already know hiragana/katakana (the phonetic parts of the Japanese writing system). If you do know those, it could help as practice for kanji and phrases.
mettamage 6y ago
I don't understand anything and I tried a bit, but IMO this is much more effort than normal. A friend of my confirmed you need to know hiragana. She has studied it a bit and could follow the lessons, whereas I couldn't (knowing next to nothing).
- hombre_fatal 6y ago
  I couldn’t even tell if the first example was showing how to create a conjunction “banana and monkey” or something else like “monkey eats banana”. Yeah, this would need a lot slower of an approach to work for day 0 beginners. Rosetta Stone and Duolingo are great examples of how it can be done.
  - toto444 6y ago
    The site is definitely not supposed to be the main resource used to learn Japanese despite being self contained. It exists to complement the learning from a more traditional app.
  - smeej 6y ago
    I didn't even get this far. I wonder if it's related to my aspie, literal brain wiring. In my head, it read like, "Symbol for monkey emoji. New symbol. Symbol for banana emoji with new symbol below it, and to the right." I certainly didn't get anywhere near "eat." I was comfortably conversant in English before I was 2, and I've learned Spanish, Greek, and Hebrew reasonably well, so it's not my language ability, or even learning with non-Latin alphabets, though I suppose it could be trying to learn without an alphabet at all? My patterns matching skills are strong. But guessing meaning of unknown symbols from their nearness to each other? Whooshes right by me!
- AnIdiotOnTheNet 6y ago
  Well the good news is that the kana are really low hanging fruit. You can learn them in a day and as long as you continue learning Japanese you will be constantly exposed to them. There are only just over a hundred of them and, unlike English, they are only ever pronounced one way[0]. Kanji are significantly more effort, with 2k characters to learn and multiple pronunciations for each. Interestingly, while you will probably hate Knaji at first and wonder why they are still used, you will come to appreciate them later as they aid reading by breaking things up[1] and disambiguating between homonyms. [0] Ok, sorry, that was the lies-to-children version. Katakana are only ever pronounced one way. Hiragana has different pronunciation for は and へ when they are used as particles. Also there's pitch accent. [1] Japanese generally doesn't use spaces.
  - imtringued 6y ago
    It took me at least a month to learn all Hiragana + katakana using getkana and they only really stuck after I began learning vocabulary. Learning them in a day is impossible. You can look them up if you want to get started on day one but that is completely different from actually knowing the by heart.
    - AnIdiotOnTheNet 6y ago
      Like I said, if you continue studying Japanese you'll be continually exposed to them which will make them stick. By "learn them in a day" I meant one day of decently focused effort to memorize the syllable for each kana character. Between flashcards, pronunciation, and writing it is very doable because that's what I did.
np_tedious 6y ago
Idk about this. I found myself translating in my head anyway
rootsudo 6y ago
Kana is really easy to learn. https://drdru.github.io/stories/1_01_food.html https://drdru.github.io/stories/1_01_food.html I found the examples are literally an image abstraction and how to read them. They're what you'd find in a similar Genki Vol 1 book, the emojis are cute. The furigana is invaluable and it helps get the structure in place, I do it similar with song lyrics from utanet.
- pchap10k 6y ago
  More or less that’s right. They’re a good step toward seeing pictographs as phonetic characters. When I first starting learning Japanese I took an intensive 7 week course. We learned Hiragana and Katakana in one week, though it took longer to really make that knowledge long term available. I forgot a lot of it after completing the course, but my first year living in Japan helped restore it all! Now my kids are learning kana and my oldest kanji. The state of the art here is still drilling and repetition. FWIW this website is cute, but without learning the characters you have to rely on verbal instruction, so as an experiment it only works after learning the Kana.
  - toto444 6y ago
    Hi OP here You definitely need to learn Kana beforehand. It used to be specified but I decided to remove it a while ago.
  - klipt 6y ago
    From what I've read, spaced repetition like Anki is state of the art for memory intensive subjects like kanji/hanzi.
    - imtringued 6y ago
      You have to mess around with plugins if you want to use Anki. Learning Kanji is also really painful. I'm still only 75% done even though it has been a year already.
keiferski 6y ago
Nice idea, but this would be much better as a slightly animated video, especially with sound effects. E.g. to emphasize the monkey eating a banana, have a sound and animation for it.
gryson 6y ago
But... why? Translation is not something to be avoided for the adult language learner. We should use every advantage at our disposal. You can spend hours trying to guess the meaning of a word through pictures and context (and not getting anywhere far), or you can spend a minute reading the translated meaning and then spend hours understanding it in context (and reinforcing it, and learning its semantics). I wonder if this idea of "avoiding the mother tongue" stems from all the misguided marketing surrounding products like Rosetta Stone which advertise "learn like children learn." Spoiler: it takes children years and years of near-constant, directed input to acquire their native language to a high degree. Edit: But I can see the value in doing this as a thought experiment ("what if translation isn't possible?"), and for that it's interesting.
- toto444 6y ago
  > But... why? I started it for two reasons. First there is close to no easy to read content when you begin learning Japanese and there is demand for such thing. The second reason is to see how far I can go like this. It is a sort of thought experiment like you say in your comment. I discovered after I started that Hans Ørberg something similar for Latin ( https://en.wikipedia.org/wiki/Hans_%C3%98rberg https://en.wikipedia.org/wiki/Hans_%C3%98rberg ). His Lingua Latina has been quite influential.
  - gryson 6y ago
    Have you tried more traditional textbooks? I understand there's a desire for an easy-to-use interactive app, but I've never come across anything remotely as good as a textbook like Minna no Nihongo (I used this book 20 years ago when I started learning Japanese - maybe I'm biased?). You can see sample pages here: https://omgjapan.com/products/minna-no-nihongo-shokyu-1-honsatsu-2nd-edition-beginner https://omgjapan.com/products/minna-no-nihongo-shokyu-1-hons... It is 100% in Japanese, but it has a companion book with translated grammar notes if needed. The majority of the book is excellent content in the form of dialogs and readings designed for beginners. I suppose the exercises work better in a classroom setting...
  - arata 6y ago
    > First there is close to no easy to read content when you begin learning Japanese and there is demand for such thing. If you know the alphabets (that is kana), you can try the easiest graded reader such as [1] and progressing from there. It assumes little to no grammar knowledge and comes with illustrations and narrated audio that you can follow along. [1] https://tadoku.org/japanese/book/4155/ https://tadoku.org/japanese/book/4155/
toto444 6y ago
Hi I am the OP. I just discovered that my submission has made it to the front page. A quick word on the rationale for not using translation. All of the classic apps (Anki, Duolingo, Memrise, ...) are to some extent translation based. Translating has three scalability issues. First, if you want to be inclusive and your product to be used by everyone you have to translate the whole product in a long list of languages. I imagine there is a ton of people who do not speak English and would love to learn Japanese ! I don't really know how to express the second limitation better than by saying 'translation is not additive' or it does not 'scale up'. What I mean by that is that if you translate at the word level (such as with Memrise), you learn words but somehow you do no learn the language (to be fair is the point of the app and that works fairly well). However a language expresses itself through sentences and translating a sentence is not the same as translating each one of its word. So you are sort of stuck at the word level and can't scale up to the sentence level by combining the low level components that words are. Third, translating sentences does not scale well due to the complexity of natural languages. If like Duolingo you base your app on translating sentences you face all the poblem faced by Duolingo. How do you translate 猿はバナナを食べる ? There are a lot of potential translations. Here are some I can think of right now : The monkey eats the banana. The monkeys eat the banana The monkey will eat the bananas. The monkeys are going to eat the bananas.
- tsimionescu 6y ago
  I think that one problem that eventually happens with this approach, which is a problem I had with Duolingo as well, is that very soon after learning a few words you need to start learning basic grammar. In these approaches that don't explain the grammar at all, you need to spend effort pattern matching to pick up the right grammatical rules to construct your own sentences, and you are never sure that you're doing the right thing. For example, in 猿はバナナを食べる, how should I understand what は and を mean? When should I use them? I can more or less easily understand that 猿 means monkey, バナナ means banana, and 食べる is related to eating. But to understand that は indicates the topic and を indicates that the word before is having an action done to them, I would have to see a LOT of examples. I think a little grammatical explanation goes a very long way in terms of the number of examples you have to show.
inetsee 6y ago
This reminds me of "English Through Pictures" by Richards and Gibson.
- bitwize 6y ago
  My sister learned early reading through the Peabody Rebus Reader method. She was given short books with titles like "Red and Blue Are On Me". A small symbol stood for each word in the early chapters, which were slowly replaced with written words until the end of the book (an example of "whole word" reading). I got strong "Red and Blue Are On Me" vibes from this page.
axiom92 6y ago
I worked briefly on a similar idea last year. The problem was to collect parallel data for translation without requiring the annotators to be bilingual. We achieved it by selecting a set of universal pictures (e.g., dog running on grass) and asking the native speakers of Hindi and English to caption it. Since the concepts in the pictures we selected were common across the languages, we were able to collect high quality data. Paper: https://arxiv.org/abs/2004.11954 https://arxiv.org/abs/2004.11954 Slides: https://madaan.github.io/res/artifacts/pml4dc-practical-data-collection.pdf https://madaan.github.io/res/artifacts/pml4dc-practical-data...
hilbert42 6y ago
I get too many of those crossed box/rectangle symbols on my Android phone to make this site worthwhile. This indicates that the Google fonts on my phone do not support those Unicode characters. Perhaps the author should consider this.
throwaway2245 6y ago
Is there no sound?
nutanc 6y ago
This is a very interesting experiment. A couple of observations from my side. 1. First goal is mentioned as, "prove that you can learn japanese without using your mother tongue". But when I click on start you show a Monkey with some Japanese text. In my mind I see the picture of a monkey and think of the word in my mother tongue, Telugu(Kothi), then I check the Japanese script. So I am not strictly learning Japanese without using my mother tongue. 2. Since Japanese script is also new to me it becomes harder. As I have to learn the alphabet first and then words and then sentences. 3. To really try this experiment out instead of the Japanese text the picture should be associated with the Japanese word and Japanese sentences. If you realize, thats how we learn language as a kid. We are shown images and told to repeat words and then sentences. 4. Deep learning tries to do something similar. Here what you are trying to do is encode the word in an image as a translation layer and then we jump across languages. Deep learning also learns embeddings and then uses those embeddings to translate. In this case, the embedding is an image, but computers can use lesser info than all the pixels to embed. 5. This embedding approach has limitations in deep learning as we can see from the current translation engines. But by using sound and image maybe our brain embeds information better and hence we learn languages better.
euske 6y ago
I've seen a plenty of "learn Japanese" sites, and always scratched my head wondering what their end goal are. The learning process should accomodate the end goals. Is that to watch Japanese anime without subtitles? To have a deep conversation with ordinary people? Or simply the language is too linguistically intriguing? (BTW, so many languages are just as intriguing as Japanese) To me, a Japanese native majoring in CompSci, learning English has obvious benefits: there are 100x more learning materials, youtube videos, source code comments, and discussions, etc. And they're better. Eventually, speaking English opens up the world. (Maybe that's also true for Mandarin today.) But Japanese? It's a niche market for a smaller number of people. I think learning Japanese is much less practical than learning English/Mandarin, so I guess you'd have to be more committed at it to actually gain something.
- stordoff 6y ago
  Personally, I want to be able to read Japanese media (mostly games) that have no English release or at best have a slightly questionable fan translation. The games I'd already playing provide _a degree_ of passive immersion, as they often only have Japanese audio and leave background elements untranslated, which is useful and motivating. It'd also be nice to some that do have an English release in their native language. Were I not a native English speaker, I suspect I'd feel the same way about English. Mandarin may be more useful/practical in some sense, but I don't come across non-English or Japanese texts often, so it's not as immediately useful to me. I'm not sure why the language is so seemingly popular, but I suspect the popularity of Japanese games and anime in the West is one of the factors.
- redis_mlc 6y ago
  > Is that to watch Japanese anime without subtitles? 20-30 years ago, Japan was an ascendant technological nation in which Americans were very interested. Japan had several dominant consumer brands and was considered an AI pioneer. But Japan lost its way when transitioning from pre-Internet hardware products to "software with hardware." So today the interest is probably about reading anime subtitles, or having a Japanese gf. I studied Japanese in SV at the Jetro-related classes, and worked in Japan briefly. https://www.jetro.go.jp/usa/topics/jetro-connects-japanese-startups-with-silicon-valley-investors-at-techmatch-019.html https://www.jetro.go.jp/usa/topics/jetro-connects-japanese-s... What's interesting is that although Japanese non-work culture is roughly opposite that of the West in many ways, it seems we share a "Protestant-like" or at least secular work ethic that is results-driven. You can see that in WW2 - the Japanese Navy was able to build itself into a world power that was briefly equal to the US Navy, and the Battle of the Coral Sea was roughly a draw. Where the cultural values differed was at the very top of leadership, and ultimately the US had aggressive, independent naval commanders, and Japan had Yamamoto (a rebel who couldn't communicate his plans effectively to traditionalists) and Admiral Nagumo (a traditional, hesitant commander who was both inflexible and unlucky - he refused to do a third wave at Pearl Harbor, and he thought the Battle of Midway was primarily intended to capture Midway!) https://en.wikipedia.org/wiki/Ch%C5%ABichi_Nagumo https://en.wikipedia.org/wiki/Ch%C5%ABichi_Nagumo
peteretep 6y ago
This is the entire premise of TOEFL
cercatrova 6y ago
This reminds me of when I was trying to learn Latin. I used a book called Lingua Latina per se Illustrata. It was entirely in Latin but it started off with simple words which you'd understand due to it being in the Latin alphabet (obviously) which Western language speakers use, and it had illustrations to show you what was happening. People could eventually read larger Latin texts from going through the Lingua Latina series of books.
princeb 6y ago
i would imagine most private language learning schools use a program where they have the first couple of classes - e.g. the basics of reading and writing the script, introductory phrases - in the student's native language before moving directly into teaching with nothing but the target language. my foreign language teacher used a lot of pictures, drawings and charades to get an idea across, if simplification into more basic words failed to do the job.
laurieg 6y ago
In my experience, adult learners strongly dislike learning without translation. If you show an adult a picture of something and say this is a “flug”, only in the very simplest cases (where they are most likely just imagining their native word for the object) will they be satisfied. If it’s a class of objects or a more abstract idea they tend to feel very unsure until they have accessed some kind of translation, even if it is a bad or misleading one. In some ways the native word is more important than the actual meaning for these learners. No native word == no feeling of understanding. A student who enjoys avoiding using translation tends to progress very well, but it’s difficult to tell if they are simply good at the target language already and don’t need translation, or their avoidance of translation is the key to their success. Teasing out the direction of causality isn’t easy. I’d love to see a language textbook produced with not language. Just pictures of situations, common items and other prompts. The idea would be to take use this textbook a long with a native speaker to listen to and learn examples of all different useful parts of the language. Even then, I feel the imagine choice would probably bias certain language features more than others.
ralferoo 6y ago
So, I'd like to try this and I think it's a viable approach... But a language, especially one with a different script that I can't read, really needs some audio clues too, because that's a big thing that'll help to be able to internalise a word. I mean I can guess banana from the picture and the repeated na sound, but I have absolutely no way of internalising the other characters. I know of the existence of katakana and hiragana, but I don't actually know any characters, and if you intend this to be teaching the language, you need some way of teaching the basic building block you're relying on. Also, as I'm learning Chinese, the kanji aren't really a problem and actually I can guess the meaning, but even then I can't internalise the kanji in Japanese with no idea what it's supposed to sound like. At best, it'll just be Chinese plus squiggles. To me it'd just be "yuan mmm banana mmm shi mmm mmm". If the goal is only to read, then I guess eventually you might start to recognise some of the grammar characters. But if the goal is to learn the language, and presumably to speak it to, then I don't see how this can succeed. Unless of course, I'm missing something obvious and there is supposed to be audio...
- toto444 6y ago
  > Unless of course, I'm missing something obvious and there is supposed to be audio... Learning Hiragana and Katakana as well as their pronunciation is a pre-requisite. That was mentioned before but for some reason I removed it.

74 points by toto444 6y ago

zrkrlc 6y ago

I feel like this will only work if you can read hiragana and katakana already.

maweki 6y ago
Not being able to pronounce or hear the words is a problem. Associating the words just with the symbols seems very taxing to the visual parts of the brain. Also using the voice/vocal/speaking parts of the brain seems advantageous.
- tsimionescu 6y ago
  Not to mention, even if you could manage to remember the symbols that way, it would still be next to useless, since you still couldn't SPEAK Japanese at all. The site creator has stated elsewhere though that the site is intended for people who already know hiragana/katakana (the phonetic parts of the Japanese writing system). If you do know those, it could help as practice for kanji and phrases.

mettamage 6y ago

I don't understand anything and I tried a bit, but IMO this is much more effort than normal. A friend of my confirmed you need to know hiragana. She has studied it a bit and could follow the lessons, whereas I couldn't (knowing next to nothing).

hombre_fatal 6y ago
I couldn’t even tell if the first example was showing how to create a conjunction “banana and monkey” or something else like “monkey eats banana”. Yeah, this would need a lot slower of an approach to work for day 0 beginners. Rosetta Stone and Duolingo are great examples of how it can be done.
- toto444 6y ago
  The site is definitely not supposed to be the main resource used to learn Japanese despite being self contained. It exists to complement the learning from a more traditional app.
- smeej 6y ago
  I didn't even get this far. I wonder if it's related to my aspie, literal brain wiring. In my head, it read like, "Symbol for monkey emoji. New symbol. Symbol for banana emoji with new symbol below it, and to the right." I certainly didn't get anywhere near "eat." I was comfortably conversant in English before I was 2, and I've learned Spanish, Greek, and Hebrew reasonably well, so it's not my language ability, or even learning with non-Latin alphabets, though I suppose it could be trying to learn without an alphabet at all? My patterns matching skills are strong. But guessing meaning of unknown symbols from their nearness to each other? Whooshes right by me!
AnIdiotOnTheNet 6y ago
Well the good news is that the kana are really low hanging fruit. You can learn them in a day and as long as you continue learning Japanese you will be constantly exposed to them. There are only just over a hundred of them and, unlike English, they are only ever pronounced one way[0]. Kanji are significantly more effort, with 2k characters to learn and multiple pronunciations for each. Interestingly, while you will probably hate Knaji at first and wonder why they are still used, you will come to appreciate them later as they aid reading by breaking things up[1] and disambiguating between homonyms. [0] Ok, sorry, that was the lies-to-children version. Katakana are only ever pronounced one way. Hiragana has different pronunciation for は and へ when they are used as particles. Also there's pitch accent. [1] Japanese generally doesn't use spaces.
- imtringued 6y ago
  It took me at least a month to learn all Hiragana + katakana using getkana and they only really stuck after I began learning vocabulary. Learning them in a day is impossible. You can look them up if you want to get started on day one but that is completely different from actually knowing the by heart.
  - AnIdiotOnTheNet 6y ago
    Like I said, if you continue studying Japanese you'll be continually exposed to them which will make them stick. By "learn them in a day" I meant one day of decently focused effort to memorize the syllable for each kana character. Between flashcards, pronunciation, and writing it is very doable because that's what I did.

np_tedious 6y ago

Idk about this. I found myself translating in my head anyway

rootsudo 6y ago

Kana is really easy to learn. https://drdru.github.io/stories/1_01_food.html https://drdru.github.io/stories/1_01_food.html I found the examples are literally an image abstraction and how to read them. They're what you'd find in a similar Genki Vol 1 book, the emojis are cute. The furigana is invaluable and it helps get the structure in place, I do it similar with song lyrics from utanet.

pchap10k 6y ago
More or less that’s right. They’re a good step toward seeing pictographs as phonetic characters. When I first starting learning Japanese I took an intensive 7 week course. We learned Hiragana and Katakana in one week, though it took longer to really make that knowledge long term available. I forgot a lot of it after completing the course, but my first year living in Japan helped restore it all! Now my kids are learning kana and my oldest kanji. The state of the art here is still drilling and repetition. FWIW this website is cute, but without learning the characters you have to rely on verbal instruction, so as an experiment it only works after learning the Kana.
- toto444 6y ago
  Hi OP here You definitely need to learn Kana beforehand. It used to be specified but I decided to remove it a while ago.
- klipt 6y ago
  From what I've read, spaced repetition like Anki is state of the art for memory intensive subjects like kanji/hanzi.
  - imtringued 6y ago
    You have to mess around with plugins if you want to use Anki. Learning Kanji is also really painful. I'm still only 75% done even though it has been a year already.

keiferski 6y ago

Nice idea, but this would be much better as a slightly animated video, especially with sound effects. E.g. to emphasize the monkey eating a banana, have a sound and animation for it.

gryson 6y ago

But... why? Translation is not something to be avoided for the adult language learner. We should use every advantage at our disposal. You can spend hours trying to guess the meaning of a word through pictures and context (and not getting anywhere far), or you can spend a minute reading the translated meaning and then spend hours understanding it in context (and reinforcing it, and learning its semantics). I wonder if this idea of "avoiding the mother tongue" stems from all the misguided marketing surrounding products like Rosetta Stone which advertise "learn like children learn." Spoiler: it takes children years and years of near-constant, directed input to acquire their native language to a high degree. Edit: But I can see the value in doing this as a thought experiment ("what if translation isn't possible?"), and for that it's interesting.

toto444 6y ago
> But... why? I started it for two reasons. First there is close to no easy to read content when you begin learning Japanese and there is demand for such thing. The second reason is to see how far I can go like this. It is a sort of thought experiment like you say in your comment. I discovered after I started that Hans Ørberg something similar for Latin ( https://en.wikipedia.org/wiki/Hans_%C3%98rberg https://en.wikipedia.org/wiki/Hans_%C3%98rberg ). His Lingua Latina has been quite influential.
- gryson 6y ago
  Have you tried more traditional textbooks? I understand there's a desire for an easy-to-use interactive app, but I've never come across anything remotely as good as a textbook like Minna no Nihongo (I used this book 20 years ago when I started learning Japanese - maybe I'm biased?). You can see sample pages here: https://omgjapan.com/products/minna-no-nihongo-shokyu-1-honsatsu-2nd-edition-beginner https://omgjapan.com/products/minna-no-nihongo-shokyu-1-hons... It is 100% in Japanese, but it has a companion book with translated grammar notes if needed. The majority of the book is excellent content in the form of dialogs and readings designed for beginners. I suppose the exercises work better in a classroom setting...
- arata 6y ago
  > First there is close to no easy to read content when you begin learning Japanese and there is demand for such thing. If you know the alphabets (that is kana), you can try the easiest graded reader such as [1] and progressing from there. It assumes little to no grammar knowledge and comes with illustrations and narrated audio that you can follow along. [1] https://tadoku.org/japanese/book/4155/ https://tadoku.org/japanese/book/4155/

toto444 6y ago

Hi I am the OP. I just discovered that my submission has made it to the front page. A quick word on the rationale for not using translation. All of the classic apps (Anki, Duolingo, Memrise, ...) are to some extent translation based. Translating has three scalability issues. First, if you want to be inclusive and your product to be used by everyone you have to translate the whole product in a long list of languages. I imagine there is a ton of people who do not speak English and would love to learn Japanese ! I don't really know how to express the second limitation better than by saying 'translation is not additive' or it does not 'scale up'. What I mean by that is that if you translate at the word level (such as with Memrise), you learn words but somehow you do no learn the language (to be fair is the point of the app and that works fairly well). However a language expresses itself through sentences and translating a sentence is not the same as translating each one of its word. So you are sort of stuck at the word level and can't scale up to the sentence level by combining the low level components that words are. Third, translating sentences does not scale well due to the complexity of natural languages. If like Duolingo you base your app on translating sentences you face all the poblem faced by Duolingo. How do you translate 猿はバナナを食べる ? There are a lot of potential translations. Here are some I can think of right now : The monkey eats the banana. The monkeys eat the banana The monkey will eat the bananas. The monkeys are going to eat the bananas.

tsimionescu 6y ago
I think that one problem that eventually happens with this approach, which is a problem I had with Duolingo as well, is that very soon after learning a few words you need to start learning basic grammar. In these approaches that don't explain the grammar at all, you need to spend effort pattern matching to pick up the right grammatical rules to construct your own sentences, and you are never sure that you're doing the right thing. For example, in 猿はバナナを食べる, how should I understand what は and を mean? When should I use them? I can more or less easily understand that 猿 means monkey, バナナ means banana, and 食べる is related to eating. But to understand that は indicates the topic and を indicates that the word before is having an action done to them, I would have to see a LOT of examples. I think a little grammatical explanation goes a very long way in terms of the number of examples you have to show.

inetsee 6y ago

This reminds me of "English Through Pictures" by Richards and Gibson.

bitwize 6y ago
My sister learned early reading through the Peabody Rebus Reader method. She was given short books with titles like "Red and Blue Are On Me". A small symbol stood for each word in the early chapters, which were slowly replaced with written words until the end of the book (an example of "whole word" reading). I got strong "Red and Blue Are On Me" vibes from this page.

axiom92 6y ago

I worked briefly on a similar idea last year. The problem was to collect parallel data for translation without requiring the annotators to be bilingual. We achieved it by selecting a set of universal pictures (e.g., dog running on grass) and asking the native speakers of Hindi and English to caption it. Since the concepts in the pictures we selected were common across the languages, we were able to collect high quality data. Paper: https://arxiv.org/abs/2004.11954 https://arxiv.org/abs/2004.11954 Slides: https://madaan.github.io/res/artifacts/pml4dc-practical-data-collection.pdf https://madaan.github.io/res/artifacts/pml4dc-practical-data...

hilbert42 6y ago

I get too many of those crossed box/rectangle symbols on my Android phone to make this site worthwhile. This indicates that the Google fonts on my phone do not support those Unicode characters. Perhaps the author should consider this.

throwaway2245 6y ago

Is there no sound?

nutanc 6y ago

This is a very interesting experiment. A couple of observations from my side. 1. First goal is mentioned as, "prove that you can learn japanese without using your mother tongue". But when I click on start you show a Monkey with some Japanese text. In my mind I see the picture of a monkey and think of the word in my mother tongue, Telugu(Kothi), then I check the Japanese script. So I am not strictly learning Japanese without using my mother tongue. 2. Since Japanese script is also new to me it becomes harder. As I have to learn the alphabet first and then words and then sentences. 3. To really try this experiment out instead of the Japanese text the picture should be associated with the Japanese word and Japanese sentences. If you realize, thats how we learn language as a kid. We are shown images and told to repeat words and then sentences. 4. Deep learning tries to do something similar. Here what you are trying to do is encode the word in an image as a translation layer and then we jump across languages. Deep learning also learns embeddings and then uses those embeddings to translate. In this case, the embedding is an image, but computers can use lesser info than all the pixels to embed. 5. This embedding approach has limitations in deep learning as we can see from the current translation engines. But by using sound and image maybe our brain embeds information better and hence we learn languages better.

euske 6y ago

I've seen a plenty of "learn Japanese" sites, and always scratched my head wondering what their end goal are. The learning process should accomodate the end goals. Is that to watch Japanese anime without subtitles? To have a deep conversation with ordinary people? Or simply the language is too linguistically intriguing? (BTW, so many languages are just as intriguing as Japanese) To me, a Japanese native majoring in CompSci, learning English has obvious benefits: there are 100x more learning materials, youtube videos, source code comments, and discussions, etc. And they're better. Eventually, speaking English opens up the world. (Maybe that's also true for Mandarin today.) But Japanese? It's a niche market for a smaller number of people. I think learning Japanese is much less practical than learning English/Mandarin, so I guess you'd have to be more committed at it to actually gain something.

stordoff 6y ago
Personally, I want to be able to read Japanese media (mostly games) that have no English release or at best have a slightly questionable fan translation. The games I'd already playing provide _a degree_ of passive immersion, as they often only have Japanese audio and leave background elements untranslated, which is useful and motivating. It'd also be nice to some that do have an English release in their native language. Were I not a native English speaker, I suspect I'd feel the same way about English. Mandarin may be more useful/practical in some sense, but I don't come across non-English or Japanese texts often, so it's not as immediately useful to me. I'm not sure why the language is so seemingly popular, but I suspect the popularity of Japanese games and anime in the West is one of the factors.
redis_mlc 6y ago
> Is that to watch Japanese anime without subtitles? 20-30 years ago, Japan was an ascendant technological nation in which Americans were very interested. Japan had several dominant consumer brands and was considered an AI pioneer. But Japan lost its way when transitioning from pre-Internet hardware products to "software with hardware." So today the interest is probably about reading anime subtitles, or having a Japanese gf. I studied Japanese in SV at the Jetro-related classes, and worked in Japan briefly. https://www.jetro.go.jp/usa/topics/jetro-connects-japanese-startups-with-silicon-valley-investors-at-techmatch-019.html https://www.jetro.go.jp/usa/topics/jetro-connects-japanese-s... What's interesting is that although Japanese non-work culture is roughly opposite that of the West in many ways, it seems we share a "Protestant-like" or at least secular work ethic that is results-driven. You can see that in WW2 - the Japanese Navy was able to build itself into a world power that was briefly equal to the US Navy, and the Battle of the Coral Sea was roughly a draw. Where the cultural values differed was at the very top of leadership, and ultimately the US had aggressive, independent naval commanders, and Japan had Yamamoto (a rebel who couldn't communicate his plans effectively to traditionalists) and Admiral Nagumo (a traditional, hesitant commander who was both inflexible and unlucky - he refused to do a third wave at Pearl Harbor, and he thought the Battle of Midway was primarily intended to capture Midway!) https://en.wikipedia.org/wiki/Ch%C5%ABichi_Nagumo https://en.wikipedia.org/wiki/Ch%C5%ABichi_Nagumo

peteretep 6y ago

This is the entire premise of TOEFL

cercatrova 6y ago

This reminds me of when I was trying to learn Latin. I used a book called Lingua Latina per se Illustrata. It was entirely in Latin but it started off with simple words which you'd understand due to it being in the Latin alphabet (obviously) which Western language speakers use, and it had illustrations to show you what was happening. People could eventually read larger Latin texts from going through the Lingua Latina series of books.

princeb 6y ago

i would imagine most private language learning schools use a program where they have the first couple of classes - e.g. the basics of reading and writing the script, introductory phrases - in the student's native language before moving directly into teaching with nothing but the target language. my foreign language teacher used a lot of pictures, drawings and charades to get an idea across, if simplification into more basic words failed to do the job.

laurieg 6y ago

In my experience, adult learners strongly dislike learning without translation. If you show an adult a picture of something and say this is a “flug”, only in the very simplest cases (where they are most likely just imagining their native word for the object) will they be satisfied. If it’s a class of objects or a more abstract idea they tend to feel very unsure until they have accessed some kind of translation, even if it is a bad or misleading one. In some ways the native word is more important than the actual meaning for these learners. No native word == no feeling of understanding. A student who enjoys avoiding using translation tends to progress very well, but it’s difficult to tell if they are simply good at the target language already and don’t need translation, or their avoidance of translation is the key to their success. Teasing out the direction of causality isn’t easy. I’d love to see a language textbook produced with not language. Just pictures of situations, common items and other prompts. The idea would be to take use this textbook a long with a native speaker to listen to and learn examples of all different useful parts of the language. Even then, I feel the imagine choice would probably bias certain language features more than others.

ralferoo 6y ago

So, I'd like to try this and I think it's a viable approach... But a language, especially one with a different script that I can't read, really needs some audio clues too, because that's a big thing that'll help to be able to internalise a word. I mean I can guess banana from the picture and the repeated na sound, but I have absolutely no way of internalising the other characters. I know of the existence of katakana and hiragana, but I don't actually know any characters, and if you intend this to be teaching the language, you need some way of teaching the basic building block you're relying on. Also, as I'm learning Chinese, the kanji aren't really a problem and actually I can guess the meaning, but even then I can't internalise the kanji in Japanese with no idea what it's supposed to sound like. At best, it'll just be Chinese plus squiggles. To me it'd just be "yuan mmm banana mmm shi mmm mmm". If the goal is only to read, then I guess eventually you might start to recognise some of the grammar characters. But if the goal is to learn the language, and presumably to speak it to, then I don't see how this can succeed. Unless of course, I'm missing something obvious and there is supposed to be audio...

toto444 6y ago
> Unless of course, I'm missing something obvious and there is supposed to be audio... Learning Hiragana and Katakana as well as their pronunciation is a pre-requisite. That was mentioned before but for some reason I removed it.