Hello everyone,
And welcome back to the most bestest German grammar course on the web.
Today, it’s time for the first big
Exercise for Passive Voice
And it’s also time for a world premiere.
Because today, I’ll show you a new type of exercise that I have created that I have never done before. And actually, I don’t think I have seen it anything like it in any of the many learning apps and tools out there either.
I have made… drumroll please…
An AI exercise assistant
*cyberpunk intensifies
Well, actually, calling it an assistant is maybe a bit of an overpromise. I know that every language learning app and their mother is rolling out AI tools at the moment, but I think the AI we have now is not good enough to really be a learning assistant or tutor, simply because it’s not reliable enough. One explanation is on point, the next one is fail.
What it is really good at though is generating sentences and that’s kind of exactly what you need when you want to practice some grammar.
And what I made is essentially that… an interactive exercise generator.
Here’s how it works
It creates a short sentence in English passive voice for you that you then translate, and it’ll give you quick feedback for your version and show you the solution.
Sounds boring enough, but the key feature is:
You can make the tool modify the example in various ways.
So you can tell it to give you the same sentence, but in past tense, or with a negation, or with a modal verb, or just longer, with more stuff.
This allows you to practice several versions of the same core sentence and really see how things change and move around. And you practice the vocabulary of the sentence at the same time.
So basically, it’s like a premade exercise with a bunch of sentences, but there’s a virtually unlimited amount of sentences for you to practice with and you can practice all these variants.
Now you might be like “Oh, nice, and I can talk to it about my mistaks to me as well and have it answer some questions.”
But unfortunately, that’s not really a good idea, because the stuff you get is just not reliable enough. Even giving feedback on your sentence and listing errors was a bit unreliable, when I was testing the tool. So I have made it so you can’t really have a free form conversation with it.
I’ll play around some more and see how the new GPT4o fares and also the other LLMs like Claude or Gemeni. But I really don’t think it’s useful to have a tool that answers questions, when 30% of what it says are just plain wrong.
You can ask it for some translations, though, if you lack some vocabulary in the example.
Anyway, I think it’s best if you just try it out.
Oh, if you want to read up on the Passive voice again, you can find the first two episodes of my module here:
And now, viel Spaß and do let me know in the comments what you think.
German Passive Practice – AI Generator
One more quick note: all the examples are about the werden-passive. So also if you see something like “was closed”, it’s about the process of getting closed, not the state.
***
***
So yeah, that’s my new tool.
It’s not perfect, but I think just the fact that you can circle through various versions of similar sentences makes it a great way to study.
I’m super curious how it went for you and how you like it, so please let me know all your thoughts and ideas and of course any glitches you ran into.
If you vibe with the tool, I will make more of them, like for example for practicing past tense, conditional, relative sentences or other grammar topics.
And actually, I do have something in the pipeline as well, but more on that soon.
Anyway, that’s it for today. Can’t wait to read your feedback, good, bad or neutral. I want it all.
I hope you enjoyed this, have a great week and I’ll see you next time.
Audio Speed
Dark Mode
Your Version: Thomas sollte gelobt wird Original: Thomas sollte gelobt werden.
8:53 am
8:53 am
Thomas sollte gelobt werden
Sorry, I didn’t get that. Please try again.
8:53 am
8:54 am
Thomas sollte gelobt werden
Sorry, I didn’t get that. Please try again.
8:54 am
8:54 am
Thomas sollte gelobt werden.
Sorry, I didn’t get that. Please try again.
8:54 am
What am I missing here?
The exercise, as it’s currently set up, is not ready for you to enter the correct version again.
You have to continue with the buttons . Any text input is not processed. I have used rather tight guardrails for this exercise.
Sometimes it gave me a German sentence to translate to German.
I tried “ The cake should be getting baked now”.
Is correct answer Der Kuchen sollte jetzt gebacken werden?
Yes, that’s correct :).
Translate this sentence:
– She should have been praised.
She could have been praised.
Your Version: Sie sollte gelobt worden.
Sie konnte gelobt worden.
Original: Sie sollte gelobt worden sein.
Sie konnte gelobt gewesen sein.
Ugh.. I’m lost :(
Me too.
What was the task and what was your version? You gave two different English versions and 4 different German ones :).
Is it possible to have a button to press when one is attempting a translation? It is discouraging to come up with a translation that is not “almost right” and have the software not even attempt a correction or even tell you the correct answer.
I’m convinced I’m not advanced enough to use this effectively.
Oh nice idea. It is theoretically possible, though I am not sure how well the system would handle it.
What I could do easily is have it give you the option to show the solution. What do you think?
This is just great. I am loving using it! It provides such good practice. I had to be brave, though! When I translated ‘food’ as ‘Lebensmittel’, I received the following response – “I’m uncertain if that was meant as a translation, but I may be confused. Can you try one more time?”. So I substituted ‘Essen’, and then all was well!
I had to give the bot very strict constraints, otherwise it goes off the rails too often. So it’s not as lenient with synonyms as it should be :).
Glad you still enjoyed it!!
I translated construction workers as “Bauarbeiter*innen” but your exercise assistant said “Bauarbeitern”. I thought I was following modern practice…
The exercise uses ChatGPT and I don’t think that has trained this new convention yet.
I earlier asked: “When do the examples get really complicated, because that’s when I make mistakes!”. You replied, “You have to select “more options””, but I can’t find “more options”…
Ah: I found “Modify Example” option…
Oh, that’s the one I meant, sorry for the confusion.
I thought this worked really well, especially the fact that you can vary the same basic text by changing tense, adding a modal verb, negative etc. That’s a really useful exercise to do. Sorry to be really picky, but as much as the German has to be accurate, so the English should be too: practice is a noun but the verb is spelt with an s – to practise.
Using “practice” for both the noun and the verb is correct in American English (and it’s becoming pretty much standard in Australian English as well).
I’m using Freedom English TM so I’ll go with “practice” then.
Maybe I’m being dim, but I haven’t been able to use any prompts — must I use some very specific wording? It gives me simple present tense sentences, but if I ask (or just use a basic command) for a different tense, modal words, etc., it gives the stock reply “I’m not certain if that was meant as a translation, but I could be mistaken. Can you please try again?” What am I doing wrong?
You cannot prompt it. All you can do is try and translate and use the buttons it offers. If you want to chat freely, you can just do it on chatgpt. This bot here is really streamlined for only one thing.
I like it, but it makes some silly comments, like “don’t add extra spaces,” when I did not, or “don’t forget to capitalize'” when I did. Odd. The instant feed back is great–keeps me engaged– and asking for a word is useful and prevents trips to the dictionary.
Does that happen often with the spaces and caps? I don’t think I can rule this out completely. There’s just a certain possibility that the “AI” connects minor corrections like this after saying “it’s all right” simply on a basis of probability. It doesn’t understand anything of what it’s doing.
I have gotten both those replies from the AI, btw.
Good to know, but I don’t think I can make this stop completely. Or if I do, it would introduce other, bigger glitches. These LLM things really have their clear limits and drawbacks.
I really like this. Vielen Dank!
Super useful. Thanks. Innovative.
I had previously asked my Chat GPT “give me 10 sentences containing X in English to translate into German and correct my translation”…and it does that, but not very well. So you read my mind, and created something that’s about the most useful of all the very very useful exercises you’ve created.
I still learn more from your long blog articles about how words are used idiomatically in German without any of these whiz bang features, but this was excellent! Thank you!!
Thanks for the great feedback :)!!
Wow, this is a lot of fun, and 100% helpful. I just motored through a few of the longer examples, consequently making some silly mistakes, and the feedback was pretty much *spot on*. This one thoroughly deserves a bookmark, and regular revisiting.
Awesome! I’m REALLY happy to hear :D!!
Wow! A terrific challenge for me. I can get the basics, but the modal examples and longer possibilities quickly test my understanding — which is great. An excellent practice tool that I plan to return to as I prepare for the B1 Zertifikatsprüfung.
Awesome, I’m happy to hear that. I’ll add a version where you get German active and change it to passive as well.
Just to make sure… you can always go back to a simple, present tense basic example by selection “start over” at the modification options. That way, you can start easy and then go to the modes you’re having trouble with with an example you already kind of know.
This was odd, my correct answer was not recognised.
Translate this sentence:
– The task must be solved by the students.
(Type your translation or or click the buttons for “help” or “new”)
Die Aufgabe muss von den Studenten gelöst werden.
I’m not entirely sure if that was intended as a translation, but I may be mistaken. Can you try again?
These glitches will happen every now and then, at least for now. I can’t really find a way to get rid of them.
Fantastic tool! I love it. I find myself constantly picking up my phone to do the translation. The exercise reinforces what I know and at the same time points out my weaknesses. Thanks a bunch.
Brilliant. Although this one escapes my understanding.
Original: Die Torte soll gestern gebacken worden sein. (The cake is said to have been baked yesterday.)
That’s not the same thing I think.
Thanks again.
It actually is.
“sollen” in a phrasing like this can express the idea of “rumor has it” or “People say”. The AI starts with a German sentence, and then translates it to English to prompt you. The AI just ended up with a confusing German example.
Hallo Emanuel! The AI got confused:
“Translate to German:
– The cake did not have to be baked by me, but not by my sister.”
And regardless of what translation I provide, it says: “I’m not certain if that was meant as a translation, but I could be mistaken. Can you please try again?”
This is bound to happen every now and then with LLM based AI. Just ask for a new example.
Hello Emanuel! This is great! Wunderbar! Louise Daoust