How to properly do L-R method, any help for my attempt appreciated.

[Le Baron]

My main objection around AI is that it can't really guide you and instead the learner is guiding the AI to guide the learner! This puts a fairly major limitation in its usefulness. The interface between technology and human use is still fairly weak.

Few people seem to want to address this.

[orlandohill]

What's the experience with this method to people who have B2 level in order to get C1-C2 level?

In my case, I'm trying to get the C1 level (through CAE Cambridge Exam) but listening part is so hard for me that I sometimes I think to give up.

L-R is useful for improving listening comprehension at any level.

Stage 2 helps to associate the written language with the spoken language. Stage 3 helps to associate the spoken language with the meaning of the words and phrases. Both help to aquire a subconscious understanding of vocabulary and grammar.

If you're having trouble perceiving and distinguishing the sounds of English, then you should probably focus on improving that first, before doing L-R and other listening practice.

[rvilcheszamora]

What's the experience with this method to people who have B2 level in order to get C1-C2 level?

In my case, I'm trying to get the C1 level (through CAE Cambridge Exam) but listening part is so hard for me that I sometimes I think to give up.

L-R is useful for improving listening comprehension at any level.

Stage 2 helps to associate the written language with the spoken language. Stage 3 helps to associate the spoken language with the meaning of the words and phrases. Both help to aquire a subconscious understanding of vocabulary and grammar.

If you're having trouble perceiving and distinguishing the sounds of English, then you should probably focus on improving that first, before doing L-R and other listening practice.

Thank you, Orlando for your time clarifying this. Just let me ask you a few things.

1- I will attend a CAE preparation given by a University, that lasts 260 hours for 10 months. IMO a lot of time, so I could think that should fill the gaps that I have in some skills to get the C1 certification (or maybe c2).

2- When you refer to "perceive and disting sounds of English" what should be a method to improve it? I thought that through an English immersion using listening could resolve it. There are other methods that I can use?

Thank you

[wherahiko]

Hi Everyone. Just a simple quick question: What's the experience with this method to people who have B2 level in order to get C1-C2 level?

In my case, I'm trying to get the C1 level (through CAE Cambridge Exam) but listening part is so hard for me that I sometimes I think to give up. It's so fustrating.

I think this method could be something new to try.

Thank you very much.

When I started with L-R in German, I was at a B2ish level in terms of the grammar education I'd had in classes. I had learnt 'all the grammar' but had not had enough exposure to comprehensible input in German. I could 'speak' haltingly, could 'write' with a lot of help from the dictionary (Krashen would have said I was a 'super monitor user'), but had real trouble understanding podcasts. At the time, I thought I just lacked vocab and/or speaking practice, and created endless Anki cards and took italki classes in an attempt to remedy this. It didn't work. Then, I L-Red Heidi, Der Herr der Ringe, then the four novels of Meine geniale Freundin. After that, I can generally understand podcasts and other native media in German. (A weird side effect is that for a while, whenever I heard someone speaking German IRL, I had the feeling they had stepped out of LOTR!) I also now have a very positive emotional relationship with German, whereas I had fear before. I hope this helps. Best wishes for your L-R journey!

[kunsttyv]

What about using these language models to create hyperliteral translations for us? Those have always been hard to come by, and I think it is a very important ingredient of L-R.

I am sure there are better prompts, but here is a first try with GPT4. It is not pretty, but that makes focusing on the audio all the more compelling. By the way I would never use English for this and would ask it to generate a translation in my native language instead.

Give me a translation of the following text, word-for-word. It is important that you use the original syntax and sentence structure. It is not important that the translation is in idiomatic English. Only respond with your translation.

Als Gregor Samsa eines Morgens aus unruhigen Träumen erwachte, fand er sich in seinem Bett zu einem ungeheueren Ungeziefer verwandelt. Er lag auf seinem panzerartig harten Rücken und sah, wenn er den Kopf ein wenig hob, seinen gewölbten, braunen, von bogenförmigen Versteifungen geteilten Bauch, auf dessen Höhe sich die Bettdecke, zum gänzlichen Niedergleiten bereit, kaum noch erhalten konnte. Seine vielen, im Vergleich zu seinem sonstigen Umfang kläglich dünnen Beine flimmerten ihm hilflos vor den Augen.

When Gregor Samsa one morning from restless dreams awoke, found he himself in his bed into an enormous insect transformed. He lay on his armor-like hard back and saw, if he the head a little lifted, his vaulted, brown, by arc-shaped stiffenings divided belly, on whose height itself the bed cover, to completely sliding down ready, hardly still maintain could. His many, in comparison to his otherwise extent pitifully thin legs fluttered to him helplessly before the eyes.

[wherahiko]

What about using these language models to create hyperliteral translations for us? Those have always been hard to come by, and I think it is a very important ingredient of L-R.

I have tried this for Maori, as there are very few resources and no interlinear texts I have been able to find.

I simply entered the following prompt into ChatGPT (free version):

Code: Select all

please give me a word-for-word interlinear translation of the following text: [text].

It gave me:

Code: Select all

Of course! Here's a word-for-word interlinear translation of the provided text:

Original Māori:
Nā, i te mea kua whānau nei a Īhu ki Pēterehema o Hūria, i ngā rā o Kīngi Herora, nā, ka haere mai ētahi Maki i te rāwhiti ki Hiruhārama, 2ka mea, “Kei hea tēnei kua whānau nei hei kīngi mō ngā Hūrai? I kite hoki mātou i tōna whetū i te rāwhiti, ā, ka tae mai nei ki te koropiko ki a ia.”

Word-for-word Interlinear Translation:
Nā, - Therefore,
i te mea - because
kua whānau nei - was born
a Īhu - Jesus
ki - in
Pēterehema - Bethlehem
o - of
Hūria, - Judea,
i ngā rā o - in the days of
Kīngi - King
Herora, - Herod,
nā, - therefore,
ka haere mai - came
ētahi - some
Maki - Magi
i te rāwhiti - from the east
ki - to
Hiruhārama, - Jerusalem,
ka mea, - saying,
“Kei hea - Where is
tēnei - this
kua whānau nei - been born
hei - as
kīngi - king
mō - for
ngā - the
Hūrai? - Jews?
I kite hoki mātou - We saw indeed
i - the
tōna - his
whetū - star
i te rāwhiti, - in the east,
ā, - and,
ka tae mai nei ki - came now to
te koropiko - the worship
ki - to
a ia.” - him.”


My next challenge is to find a way to do this at scale and with tab spaces instead of dashes so I can use it to make an interlinear text of a whole book.

It works fine for parallel translations too.

[emk]

What about using these language models to create hyperliteral translations for us? Those have always been hard to come by, and I think it is a very important ingredient of L-R.

I have tried this for Maori, as there are very few resources and no interlinear texts I have been able to find.

I have tried something related with GPT-4, with the goal of automatically explaining tricky text on flash cards. You can see an example in my log:

Can you break down "la primera salida que de su tierra hizo el ingenioso don Quijote"? I'm not sure what's going on here.

Certainly! The phrase "la primera salida que de su tierra hizo el ingenioso don Quijote" is a bit complex due to its structure, so let's break it down:

1. "la primera salida" translates to "the first departure" or "the first sally." Here, "salida" refers to the act of leaving or going out, indicating an expedition or venture outside one's usual confines.
2. "que" is a relative pronoun here, meaning "that" in English. It's used to introduce a relative clause, providing more information about "la primera salida."
3. "de su tierra" means "from his land" or "from his homeland." "Su" is a possessive adjective that corresponds to "his," and "tierra" refers to land or territory.
4. "hizo" is the past tense of the verb "hacer," which means "to do" or "to make." In this context, it translates to "made" or "undertook."
5. "el ingenioso don Quijote" translates to "the ingenious Don Quixote." "Ingenioso" describes someone who is clever or inventive, and "Don Quijote" is the name of the protagonist.

Putting it all together, "la primera salida que de su tierra hizo el ingenioso don Quijote" translates to "the first departure that the ingenious Don Quixote undertook from his homeland." This phrase indicates that the passage is about Don Quixote's first venture or journey from the place where he lived.

This isn't terrible! But I feel like if you try to force it into an interlinear model, you'll need to get clever. It will make lots of mistakes, and the harder you push it away from its training set (and from major languages), the worse it will do.

I would recommend working in small chunks (no more than a sentence at a time), giving it several examples of the notation you want, and chosing a notation that matches the popular conventions for linguistics papers about your language. You might actually want to break the sentence into tokens and use the "function calling" features at the API level. Substudy does this to get rather decent translations of subtitles (and even then I occasionally need to retry a translation because it's missing a line or something).

This is also a use case where I'd consider splurging for GPT-4-Turbo, or at least comparing the output of GPT-3.5-Turbo and GPT-4-Turbo.

Here's a half-baked attempt. But I'd want to give it maybe 5 fully-worked examples, and to use the function calling API to cut down on the verbosity—and to provide a way to check if it returned the right number of words.

Also, if you pick a difficult and obscure language with inconsistent academic conventions, like Middle Egyptian, it's going to give you non-stop hallucinations. Probably for many languages where you really need the interlinear format, it's likely to be struggling already.

I don't think, however, that this is likely to be affordable at the scale needed for L/R. Maybe you could get a few pages at a time, under ideal conditions, and with tools that called the API and post-processed the results. I've managed to get natural line-by-line translations between a few hundred subtitles for major languages. And I'm experimenting with prompts like "Can you break down '____'?" (above), which I would only trigger manually when I actually needed them. But converting (say) an entire Greek book to interlinear format probably isn't going to happen.

[orlandohill]

I will attend a CAE preparation given by a University, that lasts 260 hours for 10 months. IMO a lot of time, so I could think that should fill the gaps that I have in some skills to get the C1 certification (or maybe c2).

Are you asking whether your University course will be sufficient to reach your goals? I have no idea. Either way, the more self-study you can do, the better.

When you refer to "perceive and disting sounds of English" what should be a method to improve it?

There is a type of practical phonetic training called ear training. I'm still learning about it myself. Both the original descriptions of L-R and the book, Fluent Forever, talk about listening to audio recordings of minimal pairs, but there's more to it than that.

It's possible that you don't need this. I'm not an English teacher, and I don't know how much difficulty native Spanish speakers typically have with English phonology.

A quick search for English ear training brought up this IPA chart and associated ebook. You should do your own search for resources too.

[rvilcheszamora]

Hi Everyone. Just a simple quick question: What's the experience with this method to people who have B2 level in order to get C1-C2 level?

In my case, I'm trying to get the C1 level (through CAE Cambridge Exam) but listening part is so hard for me that I sometimes I think to give up. It's so fustrating.

I think this method could be something new to try.

Thank you very much.

When I started with L-R in German, I was at a B2ish level in terms of the grammar education I'd had in classes. I had learnt 'all the grammar' but had not had enough exposure to comprehensible input in German. I could 'speak' haltingly, could 'write' with a lot of help from the dictionary (Krashen would have said I was a 'super monitor user'), but had real trouble understanding podcasts. At the time, I thought I just lacked vocab and/or speaking practice, and created endless Anki cards and took italki classes in an attempt to remedy this. It didn't work. Then, I L-Red Heidi, Der Herr der Ringe, then the four novels of Meine geniale Freundin. After that, I can generally understand podcasts and other native media in German. (A weird side effect is that for a while, whenever I heard someone speaking German IRL, I had the feeling they had stepped out of LOTR!) I also now have a very positive emotional relationship with German, whereas I had fear before. I hope this helps. Best wishes for your L-R journey!

Thanks dude. I started today the inmersion in L-R, firstly listening in Spanish and reading in English (L1-L2). I took a book called "Heavier than Hell" a Nirvana biography to test the method. The principal challenge here (for a spanish student) is find bilingual books with audio, but I found an amazing site to create audiobooks for free, without word limits (I can share with you internally).

I have tought use the L1-R2/L2-R2 process per chapter, I mean, first book chapter in L1-R2 and when i finish the chapter, I switch to L2-R2-

I expect test the method with 3 hours per day (early in the morning). I'll start the CAE preparation by the end of March, so, I hope to arrive with the method tested.

Thanks !

[orlandohill]

I started today the inmersion in L-R, firstly listening in Spanish and reading in English (L1-L2).

Are you sure you understand the three main stages of L-R? In Stage 1, it's fine to listen to the audiobook in your native language, Spanish, instead of reading the book, but I wouldn't recommend simultaneously reading in your target language, English. You'll just reinforce pronunciation mistakes through your subvocalisation.

For Stages 2 and 3, you should only be listening in your target language, English, while reading in English or Spanish. When you get to the point where you can understand almost every word of the book in English, then you could drop Stage 3, and just learn new words in Stage 2 through context.