Hacker News Reader

Comment by celltalk | Hacker News Reader

My best friend and I had been trying to learn Dutch for a while. About a month ago, we picked up some Donald Duck books in Dutch, thinking they'd be fun and simple to start with. But soon enough, we found ourselves constantly switching between the book and our phones, struggling to follow the story—definitely not the immersive experience we had hoped for...

That’s when I remembered those books with one language on one page and the translation on the opposite page. Inspired by that concept, I thought, why not use AI to create something similar, but even more interactive?

So, we built DuoBook.

Here's how it works:

1) Start writing your story in your language.

2) Select the language you want to learn.

3) AI helps complete the story, side-by-side with your native language.

It’s still early days, and it might not be perfect, but it's genuinely helping us—and we hope it helps you too!

Check it out: duobook.co

View on HN

At least with German Donald Duck the language used there is.. special in a good way. The early translator Erika Fuchs devised a whole new variant of German that:

- uses words and phrases that are either of rare historic origin or completely made up new ones

- verb forms so uncommon that verb form she used frequently (the Infliktiv) has a second inofficial name: the Erikativ

- she frequently borrowed from the biggest writers and poets in the German language in her translations

- for the younger figures there is an entirely made up youth slang that is both appealing and incredibly entertaining to read

The english originals are utterly boring to read in comparison. Her work has a literaric and entertainer quality of the kind that made generations realize there is no real border between serious high brow literature and comics.

Hey! I'm also learning Dutch and trying it out. A couple of things I discovered. For reference, I'm a native English speaker and speak Dutch at a B2~ level but my reading vocabulary is probably closer to B1.

The "long" story is not that long. I was expecting something closer to 1000 words or more. Using your Donald Duck example, I doubt the long story was more than 4 or 5 comic book pages. I started with the magic mushroom example, intermediate difficulty, and long length.

I also generated another story using the "lost puppy" prompt, but this time advanced difficulty and long. The difficulty was ramped up which I appreciate, but the length was even shorter than before!

The speech synthesis is garbage, but I'm sure you already know that. For a free service (for now), I understand why some limitations are in place, but it doesn't look good for your product unless you're targeting beginners. I'm not sure what you're offering is useful to anyone above that level. 3 stories per day for me is about 10-15 minutes on advanced (including the generation waiting time).

I wish you luck with the project! Imo, your time is best spent now optimizing your spend so you can provide higher quality audio to go alongside the text. I'm sure you will have plenty of content generated, but surfacing that to users without it feeling icky might also be a challenge.

> That’s when I remembered those books with one language on one page and the translation on the opposite page.

One term for this is "Parallel Text" (https://en.wikipedia.org/wiki/Parallel_text).

Cool idea! I've been thinking about learning German - I'll have to give this a try.

Please do! Let us know :)

Do you have a few samples that can be read without registering?

Does it highlight the matching words?

If you go down on the main page, you will see an example of Spanish to English.

And, we don’t have language alignment yet. Still, it highlights some of the “hard” words.

You'll probably want to show examples in all the languages you target. I wanted to see how good the Vietnamese is for my kids but signing up (especially on mobile) just to see if the tool is even useful is enough friction that I didn't bother.

Sad to hear. We will think about a solution for this. But, honestly I don’t know anyone who knows Vietnamese, so it would be such a useful feedback. That’s what the internet is for… grilling random people on the internet :)

I tried at the top and never scrolled down.

Some comments about the example:

It's weird that the Spanish version is complete and the English version appears sentence by sentence. I'd prefer o see both.

Highlighting by sentence is fine. Matching word would be better, but I can imagine it's a nightmare to implement.

Why does it read aloud only the Spanish version? I'd like to hear any of them with a click.

You can change the settings for showing the translation and auto-play.

Nice job. I'd recommend "publishing" some bilingual books that have already been made on the platform as a better pattern for discovery since content creation requires signing up.

Thank you! That’s a great suggestion maybe we can show the latest generated stories by other users.

That just leads to abuse. Think about it.

Go get something out of Project Gutenberg that is popular, and not too long.

Translate it into 3-4 major languages. Give those away for free.

Even better, test your translations against actual public domain translations if they exist, and be transparent about WER for each language.

If you want this to be a business, you’re going to have to do some serious business-like things.

FWIW translating into "major languages" isn't necessarily much help. If I speak one of the supported languages, say, Vietnamese, me being able to look at the German/Chinese/French is useless because 1) I can't read those, I already said I read Vietnamese and 2) the translation models often vary widely between languages so doing well on French doesn't mean they will do well on something else.

> 2) the translation models often vary widely between languages so doing well on French doesn't mean they will do well on something else.

Especially this. I often come across AI products that claim to do well in 100+ languages, show some really good DE/FR/SP/RU examples, then I try it with my language (Slovene) and am just disappointed. If you claim to support all those languages, please have a sample result in all of them. Even if they aren't all equally good, it comes across as more genuine than making bold claims that anyone who speaks a language with < 10 million speakers knows likely aren't true.

Word Error Rates (WER), are the standardised metric for understanding that. The number is the number of words you would need to add, remove or replace per 100 words (so lower is better). Spanish these days is often down in the 2-3 region, English in the 4 region, and Finnish in the 7-8 region. Slovenian is typically 15-20, and Punjabi, Bengali and other SE Asian languages hit at 40+.

By far the biggest factor seems to be the amount of translated material in that language, followed by how "obvious" the rules of the language are making it easy to decompose algorithmically.

I think that's why calculating and showing the accuracy metrics (like WER) against known good translations, is useful. You can even highlight words that the model struggles with, giving the reader useful context.

Translation models are getting better all the time - it's a weird artefact of transformer architectures that got missed in the GenAI hype, that they're pretty great at translation, especially across languages with smaller training corpuses - but you should definitely know if the text you're reading is only likely to be 90% "correctly" translated.

I think, judging from your responses in this thread, that you're focusing too much on the 'generating stories' part. IMO that's the least attractive/useful part of your offering. The 'read something in two languages' is what use useful to your users. I made something not quite like your app but related: a tool to translate (epub) ebooks into two-column ebooks with a translated version on one side, so you're not constantly googling/chatgpting things. What you have has the potential for more interactivity though (like, my dual language ebooks can't highlight words to match from left to right or vice versa). I would love your tool if it wasn't for the 'the content is AI generated' part. I'm not looking to add more AI slop into my life, but that doesn't mean I'm against it for actually useful purposes, like translations/language learning.

This is such an amazing idea and I would pay for it right now as finding text for A1-A2 is virtually impossible if your language is not incredibly popular

How much would you be willing to pay for it for a monthly subscription?

We were thinking about a pricing page with few pro options. For instance, longer text generation, access to better LLMs, or much better TTS etc.

How long would the stories be? For the ability to generate up to 20 pages of text per month, I would pay between 10-15 dollars per month literally right now.

Can you reach out to us through the contact form? Maybe we can setup a 10-min call, because we’ve implemented a pro tier.

At first I had trouble coming up with an idea for a story. I think it could use more examples. I chose "Ayn Rand gets added to Mt Rushmore" and it made a nice story about it. The next one I might try is "Dingoes permitted as pets around the world". Basically anything you can imagine. I like the vocabulary quiz. I did Spanish and it had a new word and a couple that I'm fuzzy on, and that was useful practice.

Edit: Year of Linux on the Desktop is another.

Edit 2: For Year of Linux on the Desktop it did 2024 as that year. Might want to add the current date to the prompt and say that to have stuff imagined in the future be after that. Another thought is to have the LLM suggest a story prompt for you.

Thank you for also trying the practice mode! Super useful feedback. I will add them to the TO-DO list.

Another thing that isn't a show-stopper but is less than ideal, is that one of the vocabulary word pairs from the dingo story I did was dingo/dingo. For some things it's worthwhile to point out that they're spelled the same, like hotel/hotel, but dingo is kind of obvious. I just looked it up and it comes from indigenous language, though it's changed a bit.

Is this a real origin story or a fake one? (Obviously written by AI)

Hahaha I know it sounds fake but it is actually true.

https://imgur.com/a/nTbBgdJ

Reply on news.ycombinator.com